Modification in Functional Discourse Grammar: State of the art and issues addressed

2.1 Distinctive features and overall organization

The general characteristics of FDG have naturally led to specific choices when it comes to the organization of the model. Thus, the “function-to-form” approach is mirrored in the model’s top-down orientation, which starts with the speaker’s intention and then works its way down to articulation. In this way, Hengeveld and Mackenzie (2008, 13) observe, “FDG takes the functional approach to language to its logical extreme,” as pragmatics is taken to govern semantics, pragmatics and semantics to govern morphosyntax, and pragmatics, semantics and morphosyntax to govern phonology. The privileged role of pragmatics is further reflected in the fact that FDG does not take the clause as its basic unit of analysis, but the Discourse Act. This means that FDG can accommodate not only regular clauses, but also units larger than the clause, such as complex sentences, and units smaller than the clause, such as single phrases or words.

In order to represent all the linguistic information relevant for the formation of a linguistic expression, FDG analyses linguistic expressions at four separate, but interactive, levels of analysis, capturing all the systematically encoded pragmatic, semantic, morphosyntactic and phonological aspects of each expression. Together, these four levels form the Grammatical Component of the model (FDG proper). This component is part of a wider model of verbal communication in which it interacts with three other components: (i) a Conceptual Component, which contains the prelinguistic conceptual information (i.e. the speaker’s communicative intentions) relevant for the production of a linguistic expression and which forms the driving force behind the Grammatical Component (see also Connolly 2017); (ii) a Contextual Component, containing non-linguistic information from the immediate discourse context that affects the form of a linguistic utterance (see also Connolly 2007, 2014, Cornish 2009, Alturo et al. 2014, Hengeveld and Mackenzie 2014); and (iii) an Output Component, which turns the output of the Grammatical Component into acoustic, orthographic, or signed output (see also Seinhorst and Leufkens 2021). A general outline of the model is given in Figure 1.

Figure 1

General layout of FDG (based on Hengeveld and Mackenzie 2008, 13).

Apart from the four levels of analysis, Figure 1 also contains the different types of primitives feeding into the representations at each level. These primitives (given in boxes) can be regarded as the building blocks needed for the construction of an utterance: they are ready-for-use elements that together make up a speaker’s long-term knowledge of a specific language. Primitives come in three kinds. First, there are the structuring primitives (frames and templates), which define the possible combinations of elements at each level. These frames and templates consist of one or more slots which can be filled by lexical or grammatical information; some of these slots are provided with functions indicating the role of an element within the frame. The second set of primitives consists of the relevant linguistic elements at each level: lexemes (at the Formulation levels) and grammatical morphemes and suppletive forms (at the Encoding levels). The third set of primitives contains operators, which represent grammatical information at each of the levels, e.g. identifiability of a referent at the Interpersonal Level, and number, tense and aspect at the Representational Level.

In constructing a linguistic utterance, the speaker first selects the appropriate primitives, starting with the (interpersonal and representational) frames, followed by the lexical and grammatical elements to fill the slots within the frames. These primitives subsequently feed into the operation of Formulation, which results in representations at the higher two levels. This information then forms the input of the two operations of Encoding, where the relevant primitives for those levels are translated into representations at the Morphosyntactic and Phonological Levels.

Each of these levels of analysis consists of a number of hierarchically structured layers, representing a particular kind of linguistic unit. In the next section, the four levels and their internal structure will be discussed in greater detail.

2.2 Four levels of analysis

The Interpersonal Level is the first and highest level of representation, and it deals with pragmatic information, i.e. “all the formal aspects of a linguistic unit that reflect its role in the interaction between the Speaker and the Addressee” (Hengeveld and Mackenzie 2008, 46). The highest layer at this level is the Move (M), encompassing an entire segment of relevant discourse. Each Move is made up of one or more Discourse Acts (A) that form its (potentially complex) Head. A Discourse Act, in turn, comprises an Illocution (F), the Speech Act Participants (P_S and P_A) and a Communicated Content (C). On the Communicated Content layer the Speaker can evoke one or more Subacts of Reference (R) and/or Ascription (T). On each of these layers, slots are provided for modifiers (expressing additional lexical information) and operators (representing additional grammatical information).

The different interpersonal layers are illustrated in (1):

• (1) a) Unfortunately, the dingo ate a koala.

•   b) (M_I: (A_I: [(F_I: decl (F_I)) (P_I)_S (P_J)_A (C_I: [(T_I) (+id R_I)_Top (-id R_J)_Foc] (C_I): unfortunately (C_I))] (A_I)) (M_I))

Example (1b) shows a Move which contains a single Discourse Act. This Discourse Act is composed of a declarative (DECL) Illocution, the two Speech Act Participants and a Communicated Content. The latter is made up of an Ascriptive Subact that evokes the property ‘eat’ and two Referential Subacts that evoke the entities dingo and koala. Both Subacts of Reference are specified by an identifiability operator: +id indicates that the Speaker assumes the Addressee will be able to identify the entity evoked by R_I (and therefore triggers the use of the definite article during Morphosyntactic Encoding) and -id shows that the Speaker assumes the entity evoked by R_J will be unidentifiable for the Addressee (which then triggers the use of the indefinite article during Morphosyntactic Encoding). The two Referential Subacts also have a pragmatic function: The first is assigned the function of Topic (Top), marking the entity in question as relevant to the current discourse and linking it to information stored in the Contextual Component. The second is assigned the function of Focus (Foc), which marks this Subact as providing new or the most salient information within the Discourse Act. This results in prosodic prominence during Phonological Encoding. The Communicated Content, finally, is modified by the attitudinal adverb unfortunately, which indicates the Speaker’s subjective attitude towards the related content. Because of the interpersonal (Speaker-oriented) function of unfortunately, this adverb is lexically specified already at the Interpersonal Level. All the other lexical elements in (1a) have a descriptive function and thus appear at the next level.

The second level is the Representational Level and deals with the semantic information of a linguistic expression, i.e. those aspects of a linguistic unit that reflect “the ways in which language relates to the extra-linguistic world it describes” (Hengeveld and Mackenzie 2008, 128). Accordingly, the layers at this level capture the different linguistically relevant types (or orders) of entities with respect to this extra-linguistic world (Lyons 1977, 442–7, Hengeveld and Mackenzie 2008, 131). The highest of these layers is the Propositional Content (p), which is a mental entity type evaluable in terms of its truth. A Propositional Content consists of one or more Episodes (ep), which, in turn, are sets of States-of-Affairs (e). Episodes are considered coherent units in that they share temporal, spatial and participant specifications. A State-of-Affairs itself consists of a Configurational Property (f^c), containing typically a verbal Property (f) and one or more Individuals (x). The Individuals usually have a nominal Property (f) as their respective head and often function as the arguments within a selected frame.

An illustration is given in (2):

• (2) a) Last week the rabid dingo ate a koala.

•   b) (p_i: (past ep_i: (e_i: (f^c_i: [(f_j: eat (f_j)) (1x_i: (f_k: dingo (f_k)) (x_i): (f_l: rabid (f_l)) (x_i))_A (1x_j: (f_m: koala (f_m)) (x_j))_U] (f^c_i)) (e_i)) (ep_i): (t_i: last week (t_i)) (ep_i)) (p_i))

Sentence (2a), with its analysis on the Representational Level in (2b), consists of a Propositional Content p_i made up of a single Episode ep_i. This Episode comprises one State-of-Affairs e_i, with a Configurational Property f^c _i as its head. The Configurational Property contains the verbal Property f_j (eat) and two Individuals x_i and x_j. Each Individual has a nominal head (f_k and f_m, representing the Properties ‘dingo’ and ‘koala’, respectively). The sentence also displays two different modifiers and two types of operators. The modifiers are the temporal modifier last week (t_i), modifying the Episode layer, and the adjectival modifier rabid (f_l), modifying the first argument x_i. The operators are the past tense operator ‘past’ at the layer of the Episode and the singular operator ‘1’ at the respective layers of the two Individuals. Finally, the semantic roles of Actor (A) and Undergoer (U) are assigned to the two Individual arguments, respectively. Here, pragmatic and semantic Formulation ends and the next two levels, the Morphosyntactic Level and the Phonological Level, are part of the formal Encoding, during which no further components of meaning can be added.

The Morphosyntactic Level captures the linear properties of, on the one hand, sentences, clauses and phrases, and, on the other hand, the internal structure of complex words. The highest layer at the Morphosyntactic Level is the Linguistic Expression (Le), which may consist of one or more Clauses (Cl). Clauses typically contain one or more Phrases and Words, as well as, potentially, other Clauses. Phrases, in turn, may consist of one or more Words, other Phrases or Clauses. Finally, at the lowest layer, Words contain one or more Morphemes belonging to one of three types: Stems (which have lexical content and can be the sole element within a Word), Roots (which have lexical content and can only be used with another Stem or Root) or Affixes (which lack lexical content and can only be used in combination with a Stem or Root). Phrases, Words, Stems and Roots are classified further with respect to their head, for instance as Verbal Phrases (Vp), Nominal Phrases (Np) and Adjectival Phrases (Ap), as well as Verbal Words (Vw), Nominal Words (Nw) and Adjectival Words (Aw). Grammatical Words (Gw) tend to correspond to operators at the Formulation levels. By way of exemplification, a morphosyntactic representation of sentence (3a) is given in (3b):

• (3) a) The dingo ate a koala.

•   b) (Le_i: (Cl_i: [(Np_i: [(Gw_i: The (Gw_i)) (Nw_i: (Ns_i: dingo (Ns_i)) (Nw_i))] (Np_i)) (Vp_i: (Vw_i: eat-past (Vw_i)) (Vp_i)) (Np_j: [(Gw_j: a (Gw_j)) (Nw_j: (Ns_j: koala (Ns_j)) (Nw_j))] (Np_j))] (Cl_i)) (Le_i))

The Phonological Level is the lowest level and usually receives its input from all the other levels above it.^[1] The highest phonological layer is the Utterance (U). It contains one or more Intonational Phrases (IP), which, in turn, are made up of one or more Phonological Phrases (PP). A typical Phonological Phrase contains one or more Phonological Words (PW). Phonological Words are further broken down into Feet (F) and Syllables (S). The representation in (4b) gives a simplified phonological analysis of the example sentence (4a). This representation also contains an operator ‘f,’ which serves to indicate a falling intonation at the Intonational Phrase layer.

• (4) a) The dingo ate.

•   b) (f ip_i: [(pp_i: /ðəˈdɪŋɡəʊ/ (pp_i)) (pp_j: /eɪt/ (pp_j))] (ip_i))

3.1 Modifiers vs operators and functions

As mentioned above, FDG makes a distinction between lexical elements (functioning as heads of interpersonal or representational layers) and grammatical elements (operators and functions, providing grammatical information about a layer). Although it is acknowledged that the difference between lexical and grammatical information is not a strict one, with the former often gradually developing into the latter through processes of grammaticalization, the distinction is nevertheless regarded as crucial, as lexical and grammatical elements exhibit different formal behaviour.

In FDG, the distinction between the lexical and grammatical elements is based mainly on two criteria: modifiability and focalizability. Thus, lexemes are those (free) morphemes that allow for modification and can be assigned focus function. Elements that can neither be modified nor focalized (typically bound morphemes) are analysed as operators or functions. In addition, FDG recognizes an in-between category of elements that fulfil only one of the two criteria (Keizer 2007, Hengeveld 2017, 31). For instance, many free grammatical morphemes (e.g. demonstrative determiners, quantifiers, conjunctions, auxiliary verbs, etc.) can be focalized, but not modified. Such elements are analysed as lexical operators (Keizer 2007, Hengeveld 2017, 31) or functions (Giomi 2020, 325, 346–8), reflecting their in-between status on the lexical-grammatical continuum (see also Giomi, this issue; Olbertz, this issue; Portero Muñoz, this issue; Ten Wolde and Schwaiger, this issue).

The various options are illustrated in example (5) for the layer of the Individual (x). In (5a), we have two lexemes heading the lexical Properties (f_i) and (f_j). Both these Properties are modifiable: in example (5a) the Property ‘student’ is modified by the Property ‘diligent,’ which in turn is modified by the Property ‘extreme’ (f_k).^[2] The indefinite article a, on the other hand, is a grammatical element, triggered by the singularity operator ‘1’ (in combination with the unidentifiability (-id) operator at the Interpersonal Level). In (5b), on the other hand, the demonstrative determiner that (focalizable, but not modifiable) is represented as a lexical operator.

• (5) a) an extremely diligent student

•   (1 x_i: (f_i: student (f_i)) (x_i): (f_j: diligent (f_j): (f_k: extreme (f_k)) (f_j)) (x_i))

•   b) that student

•   (that x_i: (f_i: student (f_i)) (x_i))

3.2 Interpersonal and representational modifiers

As we have seen, modifiers can be added at both the Interpersonal Level (see example (1)) and the Representational Level (see examples (2) and (5)). Modification at these two levels is different, first and foremost, in terms of function. Interpersonal modifiers have a Speaker-oriented function, such as expressing the Speaker’s attitude towards (e.g. unfortunately, surprisingly) or emphasizing the Communicated Content (e.g. absolutely, totally), or indicating the Speaker’s manner of carrying out the Illocution (e.g. frankly, honestly), or indicating that the Communicated Content has been obtained from another source (e.g. allegedly, supposedly) (see also Kaltenböck and Keizer, this issue; Kojadinović, this issue). On the other hand, representational modifiers function as additional specifications in that they provide additional information about the extra-linguistic entities that are captured by the various semantic categories at the Representational Level. Therefore, subjective modality adverbs (e.g. hopefully) modify the Propositional Content, reality (e.g. actually) and frequency (e.g. frequently) adverbs the State-of-Affairs, manner adverbs (e.g. angerly) a verbal Property, etc. Similarly, adjectives like poor in poor Jane (expressing the Speaker’s sympathy towards a referent) and utter in an utter fool (intensifying the Property denoted by the following noun) are represented at the Interpersonal level, whereas descriptive adjectives, denoting properties of a referent in the extra-linguistic world (either objectively, e.g. ‘red’, ‘square’ or subjectively, e.g. ‘kind’, ‘awful’) are analysed at the Representational Level (see also García Velasco, this issue; Keizer, this issue; Kemp and Hengeveld, this issue).

Interpersonal and representational modifiers can, according to Hengeveld and Mackenzie (2008, 128–9), be distinguished by their differences in truth-conditionality, as shown in examples (6) and (7):

• (6) Frankly, Sheila is ill.

•   a) No. (She isn’t.)

•   b) ^*No. (You are not being frank.)^[3]

• (7) Peter told me frankly that Sheila is ill.

•   a) That’s not true. (She isn’t.)

•   b) That’s not true. (He was not being frank.)

•    (Hengeveld and Mackenzie 2008, 128–9)

In (6), the speaker can deny the whole Propositional Content (see (6a)), but not the information expressed by the interpersonal adverb frankly (see (6b)); therefore, frankly cannot be part of the Propositional Content, and is, hence, non-truth-conditional. In (7), frankly is a representational (manner) adverb and, therefore, part of the Propositional Content, so that it is possible to negate the adverb. This adverb in this context is then truth-conditional (see (7b)).

Further support for the distinction between the two groups of modifiers is found in their relative ordering within the clause or phrase. According to the top-down and outward-inward placement of linguistic units during linearization at the Morphosyntactic Level (see Hengeveld and Mackenzie 2008, 311–4, 377–80), interpersonal adverbs will most likely be placed in the (left) periphery of the clause, with representational adverbs being given more central positions; this is demonstrated in (8a). In this example, unfortunately comes before the representational (evidential) adverb presumably (Keizer 2015, 189).^[4] In the same way, in (8b), the interpersonal adjective poor (expressing the Speaker’s sympathy for the entity evoked) is placed more peripherally than the representational (descriptive) adjective innocent (Keizer 2015, 190, 220–1):

• 8) a) She unfortunately presumably saw him again last week.

•   b) a poor innocent dinosaur

3.3 Determining the scope of interpersonal and representational modifiers

Just like we need criteria to determine whether a modifier is interpersonal or representational, we also need criteria to determine which exact interpersonal or representational layer a modifier belongs to. As already mentioned in the previous section, the starting point for assigning a modifier to a particular layer is the function it performs (e.g. Keizer 2018a,b, 2020, Kemp 2018). The scope of modifiers is, however, also reflected in their formal behaviour. Thus, as in the case of interpersonal and representational modifiers, modifiers within each level also tend to appear in a relative order reflecting their scope, with modifiers of higher layers occurring in more peripheral positions than those of lower layers. An example can be found in (8a), where the temporal modifier last week, modifying the Episode, is in a more peripheral position than the frequency modifier again, which scopes over the State-of-Affairs within the Episode.

Further evidence for the scope of an interpersonal or representational modifier comes from coordination, since only modifiers scoping over the same layer tend to be coordinated. Thus, example (9a), where the two subjective modality adverbs probably and certainly are coordinated, is fully acceptable, whereas (9b), where certainly is coordinated with the factuality adverb really (scoping over the State-of-Affairs), is at best questionable (see Keizer 2018a,b, 2020):

• (9) a) Bull was always, first and foremost, a virtuoso both of technical invention and obviously of performance, even in compositions probably or certainly intended for the organ where he appears as the direct heir of Preston and Blitheman. (BYU-BNC, non-academic)^[5]

•   b) ^?*… even in compositions really and/or certainly intended for the organ where he appears as the direct heir of Preston and Blitheman.

Finally, the scope of a modifier determines whether or not it can occur in the complement of a specific verb or noun (Hengeveld and Mackenzie 2008, 363–5). For instance, since the propositional attitude verb believe in (10a) takes a Propositional Content as its complement, this complement can contain adverbs modifying this layer, such as probably, as well as all lower-layer modifiers, such as the factuality adverb really; it cannot, however, contain (prosodically integrated) higher-layer (interpersonal) adverbs like reportedly (which modifies the Communicated Content). In (10b), we find the aspectual verb continue, which takes as its complement a Configurational Property. As such, this complement can only contain modifiers that occur at the layer of the Configurational Property or lower layers, such as the manner adverb systematically; modifiers belonging to higher layers, such as probably, are, however, not allowed.

• (10) a) Eventually Angel came to believe that she probably/really/*reportedly had killed d’Urberville. (BYU-BNC, fiction; adapted)

•   b) The real irony is that if Israel continues to systematically/*probably close all community service institutions … (BYU-BNC, non-academic; adapted)