1Human language is inherently symbolic in nature (Deacon 1997; Tomasello 1999; see also Hoffmann 2017a). All the languages of the world possess thousands of words, i.e. arbitrary pairings of FORM and MEANING (cf., e.g. the German words Apfel and its Hungarian equivalent alma, which have the same underlying meaning ‘apple’, but different associated conventional forms ([ˈapfl] and [ˈɒlmɒ])). More than 30 years ago, several linguists pointed out that it is not only the word level that crucially depends on FORM-MEANING pairings. As Construction Grammarians (for an overview see Hoffmann & Trousdale 2013) argue, FORM-MEANING pairs, i.e. ‘constructions’, are actually the central unit of human language, and all linguistic levels from morphology over phraseology/idiomaticity to syntax proper can be captured by constructions. Now over the last decades, a large body of scientific evidence from first language acquisition (for an overview cf. Diessel 2013), second language acquisition (cf. Ellis 2013), psycho- as well as neurolinguistics (cf. Bencini 2013 and Pulvermüller, Shtyrov, and Cappelle 2013) has accumulated that supports this constructionist view of mental grammar.
2As we will show in the present paper, however, Construction Grammar is not only a successful syntactic framework. In the following, we will argue that larger complex units such as texts, genres, discourse patterns can also be elegantly captured in a constructional analysis, thus underlining the systematic nature of language beyond the sentence (cf. Bergs 2008; Hoffmann & Bergs 2014; Hoffmann 2015). In particular, we will point out that genres and text types can be described as complex FORM-MEANING pairings that exhibit the same entrenchment and schematization effects as morphological or syntactic constructions. Genres and text types, consequently, are mentally stored in a bottom-up fashion as conventionalized constructions provided they either have a non-compositional meaning/function or occur with sufficiently high (type and token) frequency. Finally, like all other types of constructions, genre and text type constructions are also stored in taxonomic constructional networks.
3Next, we will first discuss the central tenets of Construction Grammar that will be relevant for our analysis of genres and text types as constructions (section 2). After that, section 3 will provide an overview of ‘Text types and genres’. In section 4, we will then present our analysis of ‘Text types as constructions’. Sections 5 discusses how our approach differs from other textlinguistic and cognitive approaches. Section 6 looks at ‘Text type changes in Construction Grammar’, before section 7 sums up our main hypotheses.
4Construction Grammar approaches (CxG, as outlined, e.g., in Hoffmann & Trousdale 2013; cf. also Goldberg 2006, 2013; Hoffmann 2017a,b.) is a theory of language that assumes that language, and knowledge of language, does not comprise universal “items and rules” (as for example advocated by generativists like Pinker 1999 and criticized by Evans 2014) but rather a structured inventory of ‘constructions’ (i.e. FORM-MEANING pairings) at various levels of granularity and schematicity, and encompassing all linguistic levels, from morphemes through the lexicon (including idiomatic characterizations) to the various patterns of syntactic structure associated with corresponding semantic, pragmatic, and discourse patterns. So, for example, single words are constructions since they have a FORM side (e.g., [tri:]) and a MEANING side (the sense ‘tree’) that are conventionally associated with each other. Similarly, idioms are constructions, as these also have a FORM side (e.g., He’s [all bark and no bite]) and a MEANING side (in this case ‘full of big talk but lacking action’). The FORM side, obviously, is more complex than just a single word. And yet, like a single word, it is conventionally associated with a particular MEANING. Note that this MEANING is holistic, or non-compositional in the sense that it cannot be derived or computed from the assembly of the individual parts.
5In addition to the classification of constructions as ‘simple’ or ‘complex’ (depending on the complexity of their FORM pole), CxG also distinguishes between constructions whose FORM side is fully fixed (A leopard can’t change its spots ‘you cannot change who you are’), partially variable (morphological constructions such as un-X, e.g. unfair, untrue, unfriendly, etc.) or fully schematic (such as the Ditransitive construction FORM: SBJ1 V4 OBJ2 OBJ3 MEANING: ‘Agent1 causes Recipient2 to receive Patient3 by V4-ing’, which licences utterances such as She sent him an email., Jack passed her the salt.; or The waiter served them their dinner.; Hoffmann 2017b: 300, 303-305).
6Importantly, the notion of the construction is obviously based on the Saussurean idea of the linguistic sign, but also crucially extends the latter (for an overview cf. Croft and Cruse 2004: 258): The FORM side of a construction can contain information not only about its pronunciation but also its (morpho-)syntactic properties. Similarly, its MEANING pole can contain not only semantic information, but also pragmatic, as well as contextual properties (including constraints on its use with regard to style and register).
7But how are constructions acquired and stored by speakers? All CxG approaches assume that non-compositional meaning (as e.g. with the idioms such as all bark and no bite or KICK the bucket ‘die)) requires the mental storage of a unit as a construction. On top of that, and in line with modern usage-based approaches (cf. e.g. Bybee 2010, 2013), most CxG approaches (see Goldberg 2006) postulate that even fully compositional constructions can be stored provided that they occur with sufficient frequency. What speakers encounter in their input are specific utterances. These so-called ‘constructs’ are created by the on-line activation and combination of several constructions. A construct such as Firefighters cut the man free., e.g., originally resulted from the combination of various constructions, such as the Resultative construction (FORM: [SBJ1 V4 OBJ2 OBL3] MEANING: ‘Agent1 causes Patient2 to become State3 by V4-ing’), the Verb-Specific cut-construction, NP-constructions as well as various lexical constructions. Once hearers are exposed to a construct, they should process it drawing on the various constructions mentioned above. At the same time, however, frequency effects crucially determine whether an additional novel construction is entrenched. In usage-based approaches, there are two different types of frequency that are thought to produce different entrenchment effects:
high token frequency leads to the entrenchment of phonologically filled constructions (Langacker 1987: 59–60; Croft and Cruse 2004: 292–93). Thus, even a construct such as She refused him the kiss can be become entrenched if it is encountered often enough. High type frequency of a pattern, i.e. different lexicalizations such as She refused him the kiss, They refused her the answer, or Leeds United have refused Sky TV entry into Elland Road ... instead, due to the domain general capacity of pattern detection and schematization (Bybee 2010), lead to the storage of a more abstract construction (such as the REFUSE.Verb-Specific Ditransitive construction). Under this view, constructional taxonomies thus emerge in a bottom-up fashion. (Hoffmann 2017b: 315)
8Recently, Traugott (2008a, 2008b) has introduced a taxonomic classification of constructions that is fully compatible with such a bottom-up usage-based approach. She distinguishes between macro-constructions (at the most general and highest level of abstraction), meso-level constructions (with varying degrees of abstraction and generality), and micro-level constructions (concrete constructions with limited generality and abstractness). Micro-level constructions (such as She refused him a kiss. or They refused her the answer) can form groups, and, given enough similarity, can lead to meso-level constructions (such as the REFUSE.Verb-Specific cxn or, at a higher level of abstraction, the REFUSE.Verb-Class-Specific Ditransitive cxn). The meso-level construction thus subsumes these micro-constructions. Meso-level constructions in turn can lead to macro-level constructions (such as the Ditransitive cxn), which in turn subsume the corresponding meso-constructions (see also Traugott 2008a). This is illustrated for the Ditransitive construction network in Figure 1:
Figure 1: Partial construction network for the Ditransitive construction
Adapted (Hoffmann 2017b: 314)
9Many approaches in Construction Grammar (cf. such as Goldberg’s Cognitive Construction Grammar (Boas 2013); Croft’s Radical Construction Grammar (2013); Bergen and Chang’s Embodied Construction Grammar (2013) and Steel’s Fluid Construction Grammar (2013)) assume that constructions are actual mental representations that are learned through language use and that show dynamic susceptibility to factors such as the above discussed type and token frequency, ease of processing, Gestaltpsychology, and prototypicality. So far, as mentioned above, Construction Grammar accounts have successfully described and analysed a large number of phenomena on the level of morphosyntax, ranging from idioms to information structure. They have also dealt with language variation (dialects, styles) and language change (for a comprehensive overview, see Hoffmann & Trousdale 2013). What is still lacking, however, are systematic studies on the role of constructions in larger linguistic units such as texts and genres (cf. Bergs 2008; Hoffmann & Bergs 2014; Hoffmann 2015). This is not to deny the importance of previous Construction Grammar analyses of genre-related phenomena (cf. Antonopoulou & Nikiforidou 2011; Bender 2007; Fischer 2010, 2011; Fried 2009; Nikiforidou 2010, 2016; Östman 2005; Ruppenhofer & Michaelis 2010). Yet, as we will show, our approach differs significantly from these lines of study, which can be characterized as taking a bottom-up perspective on genre. So, while earlier research mainly focused on the internal analysis of particular constructions that might appear frequently in certain text types/genres (see for example, Nikiforidou 2010, 2016, who looks at past + now in literary genres and beyond, or Ruppenhofer and Michaelis 2010, who investigate genre-based argument omissions), we carry out an analysis which primarily takes a bottom-up and holistic perspective of genres and thus complements these previous constructionist accounts. Crucially, we argue that linguistic text types/genres themselves are construction-like objects that are acquired in a usage-based fashion just like ‘ordinary’ constructions. This paper is thus part and parcel of a major research project which aims at demonstrating that constructions in the technical sense can be seen as central units not only of language, but of human cognition generally (see also Hoffmann 2017c).
10In the following, we argue that texts and text types can be treated as constructions in a technical sense, i.e. as conventionalized form-meaning pairings, albeit of a very complex nature. The following discussion deliberately tries to avoid the pitfalls that come with discussions of the kind “what is a text?”, “what is a genre?” or “what is a text type?” (cf. Moessner 2001; Ehlich 2011; Fludernik 2000). While these questions can certainly be interesting for literary studies and text linguistics as such, especially when it comes to categorization, classification and definitions, they do not seem to be very rewarding for an approach that explores the cognitive underpinnings of how speaker/writers actually acquire different types of texts and genres, especially new ones, and how lay hearer/readers actually perceive these differences, and what the basis for their very own classification is (beyond scholarly questions such as ‘genre’ versus ‘type’). We will thus use the terms genre or text type practically interchangeably and will concentrate instead on genres/text types as micro-, meso-, and macro-constructions, as well as their corresponding features on the both the FORM and the MEANING pole. In order to illustrate and contextualize our approach, however, we need to briefly outline in the following section how it relates to more standard literary studies approaches.
11To begin with some fairly obvious everyday distinctions: for centuries, people have produced distinctly different types of texts. A broad classification of texts into epos, drama and poem, e.g., has existed since antiquity. With the advent of new forms, such as the novel in the eighteenth century, we also witness the development of new classifications. So modern textbooks would rather speak of verse, prose, and drama (cf. e.g. Nünning & Nünning 2014). Within these very abstract patterns again we see some sub-types, such as novel and short story for prose, Renaissance and absurd drama, or sonnet versus epic poem. This categorization in turn breaks down into historical novel and crime fiction, Renaissance comedy and tragedy, or Shakespearean and Spenserian sonnet etc. And we can go one step further: within the group of crime fiction we have Patricia Cornwell’s Postmortem (1990), and Shakespeare’s Much Ado about Nothing (c. 1598) and Spenser’s sonnet No. 60 (1595), can be seen as individual and concrete instances of their respective categories.
12Two observations are important here. First, this debate has focused on literary genres for a long time, dividing the field between literary scholars (dealing literary genres) and applied linguistics (dealing with everyday text types). The approach presented here does not support that distinction, but rather suggests that due to the very architecture of Construction Grammar as a psychologically plausible account of human verbal behaviour, we can develop a model that accounts for both these fields and their respective phenomena in the same way. Second, closely related to our previous point, the categorizations described above seem to rest on two somehow interconnected phenomena. On the one hand, the categorization of genres has always been a matter of academic debate and normative prescriptivism. In other words, scholars define genres and genre markers. On the other hand, even people who have received very little schooling still seem to recognize genre conventions and distinctions, perhaps through factors such as exposition and cliché, but certainly not through academic training. In other words, even somebody with very little schooling or training can (and does) easily distinguish between prose, verse, and drama, and probably even a few subgenres, such as “serious fiction” and “romance novel” (even if they might use alternative, non-canonical terms for these – note that we are aware that future experimental studies are, obviously, needed to explore these classifications made by lay readers in more detail). Needless to say, both these aspects are intricately interwoven, mutually dependent and highly dynamic (see above on the historical development of genres and genre conventions). One aspect that needs to be stressed here, though, is that genre is not an ontological category in the sense that it exists in its own right independently of our classification. So, no matter whether scholars or language users explicitly categorize linguistic products into different genres, it is the subconscious mental categorization that we are interested in. In the following, we will show that this categorization crucially depends on exposition, frequency, pattern recognition and intention reading (just like L1-acquisition; see Tomasello 2002). In essence, we offer a usage-based constructional theory of how speakers end up with genre-like information in their minds.
13We claim that texts and text types constitute constructions in a technical sense, that is complex pairings of FORM and MEANING which are stored in taxonomic networks with micro-, meso- and macro-levels. They are conventionalized FORM-MEANING pairings with either non-compositional meaning/function or with sufficiently frequent occurrence that are acquired in a bottom-up, usage-based way. Let us illustrate this claim by starting with the lowest level of abstraction: All texts (e.g. Postmortem, Much Ado about Nothing, a recipe for manicotti, the lease for your apartment etc.) obviously have a concrete (and invariable) FORM pole, which can include not only their concrete individual linguistic elements (morphemes, words, phrases, paragraphs) but also contextual features, such as book cover, shape or size, layout, their place at the bookstore, etc. Their MEANING pole will comprise rich semantic information (the ‘story’, including the evoked frames and frame elements; see also Boas 2016), but also connotational meaning (that is aesthetic and or emotional attributes such as ‘formal and solemn’ for sonnets or idiolectal judgements such as ‘our song’ for a piece of music that has individual biographical relevance for one specific couple), pragmatic information such as the overall text function (e.g. ‘representational text’ in the case of news reports or documentaries, or ‘directive text’ in the case of advertisements or recipes; see Schubert 2012: 104-105), as well as social (what kind of person produces this kind of text for which intended audience?) and contextual usage constraints (in which situations do speakers/writers produce a specific text; see also Bergs and Hoffmann 2014; Hoffmann 2015: 286-289). For a hearer/reader who first encounters a text, it will therefore be a highly complex construct. In line with usage-based approaches, we argue that input effects crucially decide whether an individual text construct is then stored in the long-term memory as a more specific text construction (Shakespeare’s sonnet 18 Shall I compare thee to a summer’s day?) or whether it gives rise to a more abstract schematic genre construction (e.g. the meso-construction Shakespearean sonnet, see below).
14Note that human memory constraints, obviously, heavily limit our ability to memorize longer stretches of text. Miller (1956) famously postulated that the working memory can only keep 7 (+/- 2) chunks in mind, and more recent psychological studies even indicate that the number is closer to 3-5 chunks (Cowan 2010). What this means is that, under normal circumstances, speakers will not store verbatim constructions of, say, the entire bible or Proust’s In Search of Lost Time. In contrast to single sentence constructs (e.g. An apple a day keeps the doctor away.), not the entire text construct will become entrenched, but only the most salient/prominent FORM-MEANING features of a specific text will be stored. Nevertheless, it is well known that professional singers and actors are capable of memorizing large repertoires of songs or even entire plays. The psychological studies on such exceptional memorizers provide important support for our constructional analysis of genres: as Noice and Noice (2006) have shown, actors do not just mechanically learn lines (that is just a string of sounds), but in order to successfully recall their parts they must imbue them with rich semantic and connotational meaning (that is, they try to identify and feel the intentions of the characters saying the lines):
Repeating items over and over, called maintenance rehearsal, is not the most effective strategy for remembering. Instead, actors engage in elaborative rehearsal, focusing on the meaning of the material and associating it with information they already know. (Seamon 2015: 45)
15Thus, actors do not just entrench FORM alone when memorizing a full play, but complex FORM-MEANING pairs, that is constructions (which are integrated by association links to the existing constructions in the construction network). This type of constructional learning of large sequences of text is not only limited to actors: Seamon, Punjabi & Busch (2010) document the case of a septuagenarian that successfully memorized all 12 books of John Milton’s epic poem Paradise Lost (and who, over a three-day period, recited all 10,565 lines). Again, this exceptional person did not just learn the FORM side of the poem but was able to fully retain and recite it over an extended period of time because he associated it with a “deep, conceptual understanding of the poem” (Seamon, Punjabi & Busch 2010: 502). Larger chunks of language can thus only be stored if they are entrenched as FORM-MEANING pairs, i.e. constructions.
16For shorter text forms, such as poems, songs or recipes the storage of an entire text construct as a substantive construction is obviously easier. Take the example of a singer:
it is clear that the singer’s knowledge of the genre, of individual songs, and of the constraints that function in them is gained without formal teaching. Exposure to variants of songs is the main learning device. At most, the novice will get the benefit of a repeated singing of the same song. From this exposure and from practice singing, the singer develops the skills needed to perform. These skills […] need not involve conscious explicit knowledge, but the ability to provide the correct response in a context. (Rubin 1995: 307)
17Rubin’s psychological analysis receives a straightforward interpretation in a usage-based Construction Grammar approach: high token frequency, that is repeated singing, of a single song leads to the entrenchment of this specific song construction. Only via ‘exposure to variants’, that is high type frequency, does a more abstract genre construction arise.
18Let us illustrate our claims using the well-known text type “Shakespearean sonnet”. As various polls by the BBC over the last decades have shown (cf., e.g., Rhys-Jones 1996 or McGough 2013), Shakespeare’s sonnet 116 (given in full in (1)) remains one of the most popular English poems:
- 1 Source: http://www.shakespeares-sonnets.com/sonnet/116
(1) Sonnet 116 (Shakespeare)1
Let me not to the marriage of true minds
Admit impediments. Love is not love
Which alters when it alteration finds,
Or bends with the remover to remove:
O, no! it is an ever-fixed mark,
That looks on tempests and is never shaken;
It is the star to every wandering bark,
Whose worth’s unknown, although his height be taken.
Love’s not Time’s fool, though rosy lips and cheeks
Within his bending sickle’s compass come;
Love alters not with his brief hours and weeks,
But bears it out even to the edge of doom.
If this be error and upon me proved,
I never writ, nor no man ever loved.
- 2 http://www.bbc.co.uk/programmes/b006qp7q
19Now some readers will, obviously, read all of Shakespeare’s 154 sonnets, thus being exposed to a high type frequency of this type of poem. Other readers, however, will only buy anthologies such Rhys-Jones’ (1996) or McGough’s (2013) or will listen to BBC radio 4’s popular programme Poetry Please.2 For the latter type of readers/hearers, (1) therefore has a high token frequency and this specific poem will, thus, be stored as a micro-construction. Yet, how detailed is this constructional representation? As we mentioned above, some people will memorize the full poem by heart, storing a complex substantive FORM pole with an associated rich semantic (and connotational) MEANING. But, for the majority of readers that only occasionally read (1), we do not claim that these automatically store the full poem. Instead, only those FORM and MEANING aspects that are salient for an individual will be stored. This, obviously begs the question which elements of a poem are predicted to be salient. Interestingly, it turns out that many of the literary devices that linguistic approaches treat as salient (that is, deviant from non-poetic language and thus prominent or foregrounded; see Leech 1969: 56-72; Stockwell 2002: 14-25, 105-119) are also the ones that psychological studies of oral traditions (Rubin 1995: 304-307) identify as supporting long-term memory storage: imagery, rhythm and poetic devices such as rhyme, alliteration or assonance. Individuals will, obviously, differ with respect to the specific constructional details that they entrench – and we are aware that our claims will have to be subjected to future experimental testing. Yet, in-line with the findings from the psychological studies discussed above, we predict that the FORM pole of the micro-construction based on (1) should include phrases with strong metaphorical imagery (e.g. Love’s not Time’s fool), passages with include the iambic pentameter (such as the opening line Let me not to the marriage of true minds (Admit impediments)) as well as rhyme pairs (minds-finds or cheeks-weeks) – all of which will be paired with semantic information on the MEANING pole. On top of that, the MEANING pole will, probably, also include information that the text function is ‘expressive’ of the emotion ‘love’. Literary scholars will, of course, rightly point out that the text function of 116 is much more complex than this and that it has to be considered in the context of all other Shakespearean sonnets (cf., e.g., Cousins 2011). As we mentioned earlier, however, we are not interested in the correct literary classifications of texts, but only intend to provide the abstractions that lay readers will be able to make on the basis of their input. And since (1) consistently appears as a ‘love poem’ in popular anthologies such as McGough’s (2013), it seems very likely that lay readers should see the main function of (1) as expressing one’s love.
20Now, once readers come across other famous Shakespearean sonnets (such as 18 Shall I compare thee to a summer’s day? or 106 When in the chronicle of wasted time), they can subconsciously generalize to a new meso-construction “Shakespearean sonnet” with the following FORM properties that consistently re-occur across the individual texts with a high type frequency:
[…] of fourteen lines, every line containing ten syllables. The first twelve do rhyme in staves of four lines by cross metre, and the last two, rhyming together, do conclude the whole (George Cascoigne 1575; cit. in Cousins 2011: 127)
21When people are asked for their knowledge of the formal elements, they might, obviously, give less concise answers than the elaborate one in the quote above (just as speakers of a language can normally not give precise formal definitions of the way they form relative clauses or tense and aspect constructions). Yet, descriptions such as “has fourteen lines”, “the first twelve rhyme ABAB, CDCD, EFEF” or “the last two rhyme” will often come up. This abstract FORM template of the Shakespearean sonnet” construction will, for lay readers, be linked to an abstract MEANING pole that identifies the function of this meso-construction as expressive of one’s love (see above). Additionally, other social and contextual features will also be associated with this construction (including a certain formality and solemnity as well as the type of situations in which people might deem it appropriate to recite it).
22Note how radically our approach differs from all other Construction Grammar approaches to genre: Recently, Nikiforidou (2016: 351) even conceded “that constructional status may be a matter of degree and that the characterization of a genre as ‘construction’ may be more appropriate for some genres than for others”. So, according to Nikiforidou, some genres might be constructions, but others are not? Does that mean one has to posit two separate mental knowledge structures (one for constructional and one for non-constructional genres)? And, more importantly, how do these abstract non-constructional genre templates emerge in the minds of language users? In contrast to other approaches, and in accordance with modern cognitive linguistics and cognitive psychology, we emphasize the prime importance of usage for the emergence of all linguistic knowledge: Genre-based knowledge is not innate. Instead, language users are exposed to complex FORM-MEANING text constructs and, guided by domain-general cognitive principles (particularly entrenchment based on type and token frequency, salience/prominence and processing constraints of the working memory), store genre constructions of various levels of abstraction in their long-term memory. Knowledge about genres (about their dynamic use as well as their formal properties) is always a pairing of FORM and MEANING information. Therefore, we put forward the hypothesis that the construction – the basic unit of human symbolic communication – is also the basic mental unit for genre information. This, consequently, is also what distinguishes our approach from other cognitive ones like Steen‘s (2011). Steen goes to great lengths to characterize genres as abstract cognitive knowledge schemas, but fails to outline how speakers actually acquire these. He carefully details many of the contextual, textual and code-related properties that play a role in genres (2011: 31-34), but his top-down analysis does not take into account how a single token of use shapes and affects the mental constructicon of a speaker. We, instead, argue that, dependent on input frequency, speakers store genre constructions in a bottom-up way and always in the same semiotic knowledge structure as all other instances of language – as constructions (and not just abstract schemas for, e.g., MEANING alone as suggested by Stockwell 2002: 78-79).
23Let us illustrate our claims by looking at another example: turning away from high art, we see that the bottom-up constructional acquisition also underlies other, more mundane genres: Just take a look at the text type recipe in (2). Due to our exposure to a high type frequency of recipes, we have entrenched a meso-construction that intuitively guides our expectations. Thus, we expect and recognize the following prototypical features of recipes in (2): that it comprises a name for the dish, the list of ingredients (for a given number of servings), the instructions for preparing the dish (in chronological and logical order), maybe the preparation and total cooking time, and maybe a picture.
(2) Manicotti Original Recipe Makes 4 Servings
Ingredients
1 pound of ground beef
1 large onion, chopped, 1 clove of garlic, chopped
1 pint ricotta cheese
8 ounces shredded mozzarella cheese
3/4 cup grated Parmesan cheese
2 eggs
2-3 teaspoons fresh parsley
2-3 teaspoons dry Italian herbs
salt to taste
ground black pepper to taste
self-made spaghetti sauce (sugo) to taste (We recommend 16 ounces)
Ample manicotti pasta (alternatively, use very thin pancakes, as in the original recipe)
Directions
Cook manicotti in boiling water until done. Drain, and rinse with cold water. Note that you can skip this step, depending on the type of pasta you use (or pancakes, of course). Using dry pasta makes the stuffing a lot easier and the pasta more al dente after baking.
Preheat oven to 350 degrees F (175 degrees C).
Stir-fry beef in a pan with one chopped-up onion and garlic until it’s fully cooked
In a large bowl, combine beef, ricotta, mozzarella, and 1/2 cup Parmesan, eggs, parsley, herbs, and salt and pepper. Mix well.
Pour 1/2 cup sauce into a large inch baking dish. Fill each manicotti shell with cheese mixture, and arrange over sauce . Pour remaining sauce over top, and sprinkle with remaining Parmesan cheese.
Bake 45 minutes, or until nice and brown and bubbly.
PREP
45 mins
COOK
45 mins
READY IN
1 hr 30 mins
24Östman (2005: 133) already claimed that speakers might “have an image of the prototypical shape of a recipe”, but does not provide details on how speakers entrench such an abstract constructional schema. Clearly, speakers differ as to how many different types and tokens of recipe constructs they encounter. Some speakers might only know the recipe for cooking an egg, while amateur chefs will have a wide range of recipes that they can cook. Just like Shakespeare’s sonnet 116, an individual text like the recipe in (2) is a concrete construct that language users will be exposed to. Speakers who only see this recipe once will only entrench a Manicotti construction with some salient FORM and MEANING properties (e.g. the name of the dish, some ingredients and the time needed for preparation – if the construction remains entrenched at this shallow level, parts or even the whole recipe will, however, be forgotten in time if they receive no further input stimulation). If they end up cooking and enjoying the dish, however, they will consult the recipe more often, leading to a higher token frequency and a detailed and strong storage of the Manicotti construction, not unlike concrete complex constructions such as idioms (see above). Given a wider interest in cooking and, consequently, greater exposure to, e.g., more pasta recipes, a more abstract pasta meso-construction will emerge in amateur chefs. Alternatively, with more and more cookbooks (for adults and children) on the market, future chefs will also see the pattern, i.e. learn the construction of recipes in general and may expect to find a certain form (and function) in cookbooks they read.
25Note that increased input frequency also accounts for the fact that certain features will be unspecified for a particular text (a recipe will sometimes come with or without pictures, it maybe online or on paper…). The meaning or function of a particular text in turn is certainly dependent on, but not solely determined by its (linguistic) form. One and the same text may be read as fiction or fact, for example. The genre of ‘faction’ exemplifies this nicely. Truman Capote’s In Cold Blood (1965) is based on fact, but can also be read as fiction (if, for example, the reader does not know about its factual background). Similarly, the linguistic form of a recipe does not per se constitute a recipe, while the function of ‘recipe’ may also be fulfilled by other forms. Note that this polystructural relationship is neither new nor surprizing from a linguistic viewpoint. The loose form-function fit in English, for example, explains how the syntactic form of imperative can also be associated with expressive functions, while the function of directives can also be fulfilled by syntactic declaratives, imperatives, and interrogatives.
- 3 Note that to some extent this parallels the distinction between coding and inferencing in contempor (...)
26The function of literary texts usually extends beyond their simple propositional content, but it is not dependent on any single formal feature. So, while a Shakespearean sonnet may basically be interpreted as woeing the addressee, i.e. to say “I love you” (see above), its very specific form and contextual features transcend this simple proposition. We thus take (literary) texts to be like signs in sense of Bühler’s seminal and foundational (1934) work3. Bühler (1934) describes the linguistic sign as consisting of a form and a meaning side, which mostly overlap. However, there are also areas where form transcends meaning, and where meaning transcends form. So there are aspects in the form that do not directly contribute to the actual meaning (e.g., whether the book is a hardcover or a paperback may not be important for its classification as novel, but is still part of its form – though a hardcover book might in some cultures rather be associated with expensive, high quality literature, of course). On the other hand, we also find meaning aspects that go beyond the actual form (e.g., if a text provokes fear, or loathing, or disgust in its readers without actually describing anything that has to do with these sentiments). Note that this also corresponds with Eco’s idea of the open work of art. Works of art do have functions, or meanings, but these are neither monolithic in the sense that there is only one ‘correct’ interpretation, nor wildly pluralistic in an anything goes manner. Their form and their specific context constrain the meanings or functions we attach to them (so that Langland’s Piers Plowman – which was composed in the late fourteenth century – can indeed by read as a criticism of society, but not as a reply to Marx).
- 4 We thank an anonymous reviewer for pointing out this line of research to us.
27This leads us to conclude that texts can actually have non-compositional functions or meanings. While the function of many non-literary genres is a lot more specific and clear (a recipe is supposed to tell you how to cook a certain dish, a lease for an apartment is a legally binding contract, the weather report is the weather report…), literary genres are more complex in this regard. Their function can be a lot broader and less specific than the meaning/function attached to traditional constructions (such as words, idioms, or syntactic configurations), but, nevertheless, it seems plausible to assume that even literary texts as a whole have meaning (at least for their individual readers) and that these meanings at least in part depend on their form. So, for example, the reader of a given novel will attach some kind of meaning to the novel as a whole. This meaning, obviously, will be mostly non-semantic, and rather vague, associative and dynamic, including aesthetic evaluation and personal biographical relevance. On top of that, genre constructions play an important role as recurrent meaning configurations that enact the social practices of a culture (Martin 1992; Martin and Rose 2008)4. We assume that the very form of the specific novel (for example its length, structure, narrative viewpoint, its cover, even contextual features such as the time and place when it was read) will contribute to the generation of this particular meaning. A change in form will necessarily result in a different meaning. This can be easily shown, for example, when readers are presented with summaries of novels they have read. The summary may be semantically very similar to the full text, and yet it does not mean the same for the readers. In the following we will tease apart the role of genre and genre conventions in these processes and we will also have to come back to other text types and phenomena which rather rest on high frequency with regard to their status as constructions.
28But before we do that, let us come back to a possible constructional analysis for texts. Concrete text constructs form the most specific (and complex) category of our constructional taxonomy, and constitute micro-constructions at the lowest level of abstraction (see Fig.1). Every text will thus become a micro-construction with a particular form that is conventionally associated with a particular meaning (with the caveats concerning memory limitations and forgetting outlined above). So, for example, Patricia Cornwell’s Postmortem (1990) counts as a micro-construction, just as Amanda Kyle Williams’s Stranger in the Room (2012). And so does the recipe for manicotti, and the one for spaghetti norma, and the one for bean soup.
29When these micro-constructions group together, they may form meso-construction text types like ‘recipe’ or ‘crime fiction’ (which are more abstract and complex). A text type (or genre) at this level does not have one single concrete manifestation (like a micro-construction does) but should be seen as a prototypical arrangement of features that belong to this particular text type. These again could range from purely linguistic markers to content and contextual features (e.g. good versus bad, the crime, a mysterious cover, prose text). And while it may be difficult to say what particular kind of function or meaning a particular genre (such as crime fiction) can have, it seems intuitively plausible that it must have a particular function (or meaning) that distinguishes it from other genres (such as historical novels or sci-fi at this particular level).
30The same applies then to the different meso-constructions, i.e. meso text-types that can be grouped together to form particular macro-constructions at the most abstract level. Note that these, of course, are usually very general and comprise only few basic features on both their form and their meaning side. Like with other meso-constructions, these features and feature lists show prototype effects, i.e. while poems are prototypically arranged by verses and lines and show rhyme and meter, there are certain exceptions, like Haikus or free verse, or prose poems. We therefore expect group membership in a particular meso- or macro-type to be guided by prototype effects and gestalt principles. In other words, speakers can categorize micro-constructions into meso- or macro-constructions given sufficient similarity on the form (and also function side). With regard to the function side of macro-constructions we can refer to the more general discussions about language functions (e.g. Jakobson 1960 or Bühler 1934) and the respective treatment in text-linguistic analyses (e.g. Fludernik 2000 and references therein, particular Werlich 1975). Whether prose texts then rather have a descriptive, or assertive, or narrative function is beyond the scope of this paper and not relevant to its core arguments.
31The framework presented here complements approaches such as Werlich (1975), story grammars (e.g. Rumelhart, Thorndyke, Mandlr and Johnson, Stein and Glen; for an overview cf. Black and Wilensky 1979) as well as several others developed in text-linguistics since the 1970s (cf. Fludernik 2000; Ehlich 2011). Werlich, for instance, posits five text types (description, narration, exposition, argumentation and instruction) that manifest themselves on three different levels: the ideal type (a kind of abstract prototype which only exists in the minds of speakers), the text form (the specific text type) and the text idiom (the concrete realization in one single text). At first sight, this sort of typology looks very similar to the one presented here. However, there are some important differences. First, Werlich posits a priori super categories (description, narration etc.) as universal categories on the basis of general functions of language – something that we do not subscribe to. Rather, macro-constructions do not exist per se but are based on actual language use out of which more general patterns (meso- and macro-constructions) are generalized. Furthermore, there is no limit on the number of meso- or macro-constructions (even though we only mentioned prose, poetry and drama in the previous section). Depending on how certain texts and text types are analysed, other categories are equally possible; in general constructional taxonomies all levels are fundamentally dynamic, as we pointed out above. Moreover, in our framework, micro-constructions (which resemble Werlich’s text idioms) need not be realizations of macro-constructions. Micro-constructions can also exist as such and in relation to the other constructions. Finally, as has been pointed out above, the development and structure of text types as constructions is based on learning and use which means that these are susceptible to frequency and prototype/gestalt effects. Instead of positing “universally pre-existing” text types as given, a constructional approach would assume that the more frequently we encounter (in language learning and language use) a given type (or more generally: pattern) the more likely this is to become more deeply entrenched and to acquire prototypical status. This can explain, for example, why the average speaker tends to expect poems to have verse, meter, and rhyme (as these, due to psychological salience and type frequency, are perceived as the conventional features of poems). This can be illustrated very nicely by looking at poems composed by average speakers for special occasions. These are frequently printed in newspapers and can be found online (e.g. http://www.familyfriendpoems.com/poem/my-dad-4):
(3) To my dad on his 50th birthday
When I was young you were always there.
To hold my hand and say I care.
Encourage me along the way
along each step a different day.
Laugh a hell of a lot and
say don’t ever smoke pot.
You teach me how to do only
good things in life
And to never judge and
only do what is right.
My dads hands were never clean
Although he was a working man he was never that mean.
He cracks half a smile only once in a while
doesn’t like to show his teeth unless they are shiny, polished, and in style.
My dad is the kind of man that performs the unexplainable
People usually stare but my dad says that doesn’t bother me because I don’t care.
There are characteristics in every person that brings out your smile
Therefore my dad has too many and I will love him a lot for a while.
Happy 50th birthday day dad
I love you.
(Source: My Dad, Birthday Poem http://www.familyfriendpoems.com/poem/my-dad-4#ixzz2VWj9yFUR)
32This poem shows some interesting features. It uses some rhymes, or at least attempts to do so (there-care, way-day, lot-pot, clean-mean, while-style etc.) usually in AABBCC etc. patterning. There is also some rhythmical organization approximating iambic tetrameter:
(4) When I was young you were always there.
To hold my hand and say I care.
33Note that while both rhyme scheme and iambic tetrameter are not used throughout (we are tempted to think that in some cases the use of these could have been to difficult for the lay author), the intention to do so is – at least for us – obvious in this case.
- 5 An anonymous reviewer correctly observes that such a claim would need empirical support. While givi (...)
34The sheer frequency and popularity of poems composed in this generic style (with frequent use of rhyming couplets and a tendency towards iambic tri- and tetrameter)5 suggests that language users have abstracted a more general pattern and prototypical features for the text type “poem” in the data they have encountered (in ballads, pop songs, children’s books and possibly also at school). Speakers thus seem to have entrenched a formal template (with the observed rhyming and meter patterns – which are psychologically salient formal devices that independently have been shown to be mentally stored by psychological studies of oral traditions (Rubin 1995: 304-307)) that for them is associated with a formal meaning/function (whose ascribed aesthetic qualities make it a construction that is chosen for particular, special situations, such as celebrations.). The development of abstract patterns out of concrete data is in line with recent accounts of language acquisition from a constructional point of view (Tomasello 2002). Once language users have internalized these more abstract patterns or features of a particular text type they can either simply copy this or, alternatively, begin to change it by adding new features (i.e. merging it with other constructions). Hoffmann & Bergs (2014; see also Hoffmann 2015) demonstrate this in their exemplary analysis of the development of football chants. These start out with complex specific chants (as micro-constructions), e.g. Are you England in disguise to the melody of the hymn Bread of Heaven. This particular chant is conventionally associated with insulting or discouraging the opponent’s team. From this micro-construction speakers develop new micro-constructions by opening up a variable slot (thus schematizing the construction) and changing single elements: Are you Derby in disguise? Are you Ireland in disguise? … The result is a meso-construction Are you X in disguise associated with the function just mentioned. At a later stage, the meso-construction is further schematized until only the melody is left intact and filled with various context dependent linguistic structures: You’re not ringing anymore! You’re not signing anymore! Does your butler know you’re here? … (see Hoffmann & Bergs 2014). The function however remains constant: supporting for the own team, ridiculing the other. Thus, a meso-construction is gradually turned into a macro-construction (see Fig. 2 below).
Figure 2 : Part of the micro- and meso-construction network of ‘Cwm Rhondda’-based constructions
Adapted (Hoffmann 2015: 284)
- 6 Shaw (2010) e.g. lists 14 different types of meso-constructions based on ‘Cwm Rhondda’, seven of (...)
35…6
36Are you England Are you Derby … You’re not singing You’re not signing …
… … … … … …
37In the previous section we have discussed the evolution of text types from micro- to macro-construction through a process such as schematization. In the following, we discuss a second process involved in the development of text types: coercion (Michaelis 2005; Boas 2011; Lauwers and Willems 2011). Coercion essentially refers to the resolution of a mismatch between form and function. In Construction Grammar this was discussed with regard to utterances like I have had too much book today where the countable NP construction book merges with the uncountable quantifier construction much. Normally, this should result in nonsense. However, the mechanism of coercion leads to a reinterpretation of the two incompatible, mismatched elements, for example through pragmatic or contextual inferencing. Thus, instead of ignoring this utterance as nonsense speakers are encouraged to read this as ‘an unspecified but large number of written materials’. The same principle can be applied to text types as constructions.
38Wilfred Owen wrote his Anthem for a doomed youth in 1917 amid the atrocities of the First World War:
(5) What passing-bells for these who die as cattle
— Only the monstrous anger of the guns.
Only the stuttering rifles’ rapid rattle
Can patter out their hasty orisons.
No mockeries now for them; no prayers nor bells;
Nor any voice of mourning save the choirs,—
The shrill, demented choirs of wailing shells;
And bugles calling for them from sad shires.
What candles may be held to speed them all?
Not in the hands of boys, but in their eyes
Shall shine the holy glimmers of goodbyes.
The pallor of girls’ brows shall be their pall;
Their flowers the tenderness of patient minds,
And each slow dusk a drawing-down of blinds.
39(5) exhibits the form of a Petrarchan sonnet combined with an English rhyme scheme: it has two parts, one octet and one sestet. The first part has ABABCDCD, the second ABBACC. In some sense, this poem can be read as the provocative merger of amorous Petrarchan/Shakespearean sonnets (form) and lament (function). In other words, readers that associate sonnets (be they Petrarchan or Shakespearean) with amorous, merry sentiments may be shocked at readings Owen’s bleak and gruesome account of the atrocities of war in that very form. But instead of discarding this as nonsense, this apparent form-function mismatch can be seen as a source of creative new interpretations and provocations. It shifts the boundaries of text type constructions. Similarly, a postmodern novel like House of Leaves by Mark Danielewski (published in 2000) creates new meaning by combining a rather well-known prototypical function (novel) with formal elements like scientific report (e.g. footnotes, commentaries, etc.). And again, we do not discard this as nonsense, but try to make sense out of this mismatch. A new meaning is coerced out of this particular combination (and the same applies, of course, to modern writers like Joyce, who transgressed the boundaries of the genre ‘novel’ with Ulysses and Finnegan’s Wake). This particular kind of mismatch and coercion can be regarded as a very important source for poetic effects, as certain approaches to art and aesthetics assume that the marked and unexpected can be a central aspect in the appreciation of art, music, literature, and the like.
40This paper explored the possibilities of couching text typology and genre analysis in Construction Grammar. In doing so, we provided a cognitively and psychologically plausible account for how speakers develop and perceive different text types/genres. Moreover, as Construction Grammar has so far predominantly been concerned with the word, clause and sentence level, this is one of the first papers to explore in how far this model of language and language processing can be fruitfully applied to higher linguistic levels such as texts and text types.
41This paper argued that genres/text types can indeed be modelled as constructions in the technical sense, i.e. as conventionalized linguistic units with both a form and a function side. These textual constructions actually occur on different levels of abstraction and granularity. From very concrete single texts to large, abstract genre categories. The development and perception of these abstract textual meso- and macro-constructions and their corresponding micro-constructions is perfectly in line with the learning and processing mechanisms uncovered in Usage-based Construction Grammar. Moreover, the special focus on concrete language use and frequency also allows us to elegantly model changes in genre categories. The model developed in this paper applies not only to everyday text types/genres such as recipes or short messages on mobile phones but can also be extended to rather rare genres such as football chants (Hoffmann & Bergs 2013; Hoffmann 2015), or even poetic genres and their very special constraints and effects. All these different categories can be described and analysed using one and the same inventory of tools, i.e. Construction Grammar. Construction Grammar explicitly allows for the inclusion of linguistic information, discourse information and cognitive factors into constructions, without giving priority to any of these factors. The information input for the formation of constructions in the minds of speaker/hearers is, by definition, rich, as Bybee points out:
This information consists of phonetic detail, including redundant and variable features, the lexical items and constructions used, the meaning, inferences made from this meaning and from the context, and properties of the social, physical and linguistic context. (Bybee: 2010: 14)
42What this means is that text types/genres as constructions can be learned, stored, and processed as linguistic, discourse-based, cognitively-complex units at various levels of abstraction. They are not fixed, but, like all constructions, are subject to constant dynamic adjustments and changes as speaker/hearers encounter new input for their constructional network. The idea of what constitutes a poem, e.g., changes when speakers encounter free verse (without rhyme or meter) for the first time.
43The advantage of employing Construction Grammar ideas for the discussion of genre or text type lies in the fact that it makes no a priori assumptions about what counts and what does not count for the recognition of a genre. All that matters is the actual data that speaker/hearers encounter, and these data may be very rich in some cases. In that sense, the approach presented here goes back to some of the general ideas developed in text linguistics in the 1970s and 1980s, but also goes beyond these in that it suggests that speakers/hearers recognise and define text types as prototypes and with a much broader understanding of what constitutes a certain type. Construction grammar can accommodate these ideas and can integrate important findings from linguistics, discourse analysis and cognitive science, which have hitherto been largely unconnected.
44At the same time, extending the scope of Construction Grammar beyond single sentences to text types and genres also pushes the boundaries of the theory from a syntactic theory to a theory of language. The next step for future research would then be to explore the degree to which Construction Grammar might actually be a theoretical model for Cognitive Semiotics, i.e. the study of how humans cognitively produce and process linguistic as well as non-linguistic meaning (cf. e.g. Brandt and Simonsen 2007: 5–6; Hoffmann 2017c).