The flesh and the bones of cohesive devices: towards a comprehensive model

4.1. The principle of structure

Generally, structure is the concept that refers to the syntagmatic order of linguistic constituents (Fromkin, Rodman, and Hyams 2007). According to SFG, a syntagm is a mere “organic configuration of elements” that gives very little about meaning (Halliday and Matthiessen 2014, 39). The following is an example of a syntagm and how it works as far as parts and functions are concerned:

According to SFG, a syntagm is important because it presents an organic configuration in terms of grammatical classes and functions. Superficially, what SFG proposes about structure and its function does not differ substantially from what other grammar theories suggest. For example, Transformational-Generative Grammar (TGG) also identifies the organic elements of syntagms via phrase structural rules (Chomsky 1957). The grammatical classes of A car hit the man would be represented in the following way:

SFG and TGG also agree in that the layers of a syntagm are organized by the relationship “is part of”. In this respect, a morpheme is part of a word; a word is part of a phrase; a phrase is part of a clause.

Emphasis on syntagms in linguistic analysis was shared by other grammar theories, such as Word Grammar (WG), which holds that information about and dependencies between individual words should be the basic component of any structural analysis (Hudson 2007). Despite the similarity between SFG, TGG, and WG in acknowledging the significance of structure, these theories are very different in the way they look at the function of structure. While TGG and WG propose that linguistic analysis should not exceed the syntagm, SFG does not consider structure as the core of linguistic analysis and suggests that analysis should transcend the sentence and consider the “system” (Halliday and Matthiessen 2014).

4.2. The principle of system

The principle of system can be considered the hallmark of SFG. The theory defines system as “the paradigmatic ordering in language” (Halliday and Matthiessen 2014, 22). Unlike structure, system involves ordering at the vertical axis rather than the horizontal. What matters in system is what could go instead of what, compared to what goes together with what, the principal ordering pattern of structure (Martin 2004). Holding the relation of what could go instead of what, system is about choices made in language and is one aspect of the meaning potential of language (Halliday and Hasan 1976; Halliday and Matthiessen 2014; Menfredi 2011). As noted earlier, SFG holds that linguistic analysis must go beyond the sentence (Halliday and Hasan 1976; Johnstone 2002; Jordan 2004; Thompson and Klerk 2002). Following from this, text and its evolvement from one clause to another is one of the main foci of SFG (Gee and Handford 2011).

Although it is occasionally, but not necessarily rightly, claimed that linguistic analysis done under the umbrella of SFG has been predominantly syntagmatic (Bateman 2008), SFG maintains that both syntagmatic and paradigmatic relations are important (Martin 2014). Halliday (2009) stresses that considering paradigmatic relations “does not mean that system is regarded as more important than structure…; it means that system is taken as the more abstract category, with structure as deriving from it” (64). At odds with SFG in this regard are a number of grammar theories which consider the sentence as the major unit – sometimes even the largest constituent (Greenbaum and Nelson 2002; Jackendoff 2002) – of linguistic analysis, and that linguistic analysis should stop there. TGG, for example, asserts that formal analysis is not possible beyond the sentence level (Coulthard 2014). In essence, this syntagm-and/or-paradigm variation stems from a deeper theoretical divide between syntax-only theories, represented by structure-oriented analysis of language on the one hand, and semantics-driven theories represented by structure- and system-based linguistic analysis on the other hand. To illustrate, theories that are driven by syntax, e.g., TGG, focus on structural, or syntagmatic configurations of language as the sole core of linguistic analysis (Carnie 2014; Hall 2005). At the centre of TGG lies a fundamental principle: “The notion ‘grammatical’ cannot be identified with ‘meaningful’ or ‘significant’ in any semantic level… [and] any search for a semantically based definition of ‘grammaticalness’ will be futile” (Chomsky 1957, 15). Although Chomsky’s 1981 Government and Binding (GB) Theory addressed lexical items as the atomic units of syntax (Black 1999), syntax was still the focus of linguistic analysis. SFG, which considers both syntagmatic and paradigmatic relations in texts, is driven by semantics (Stubbs 2014), and, therefore, links grammar to meaning making as configured through systems and networks of horizontal and vertical relations among various text elements. Figure 1 summarizes the above discussion about the syntagm–paradigm theoretical divide.

Figure 1.

Systemic vs. structural linguistic analysis.

At the borderline of the syntagm/paradigm divide is the Applicative Universal Grammar Theory, which defines sentence structure as the network of syntagmatic and paradigmatic relations between sentence parts and all other expressions that can be substituted for these parts (Shaumyan 1987; Shaumyan and Segond 1994). This theoretical stand has found its way to the Arabic context as syntax and semantics were occasionally described in terms of structure and word order (Bahloul 2008; Holes 2004). This view of structure as encompassing both horizontal and vertical relations is an oversimplification of the broad divide between syntagm and paradigm on the one hand, and the underpinning distinction between syntactic orientation and semantic orientation to language on the other hand. SFG makes clear distinction between structure (sentence level) and system (text level) (Gee and Handford 2011). In order to put this within its wider context in the theory, SFG introduced the third principle, which is stratification.

4.3. The principle of stratification

According to SFG, “language can be explained as a multiple coding system comprising three levels of coding, or strata” (Halliday and Hasan 1976). The three strata are (1) semantics, which is realized by (2) the lexico-grammar, which is realized by (3) sounding/writing. Figure 2 below outlines the three strata according to SFG as adapted from Halliday and Hasan (1976).

Figure 2.

The three strata of language according to SFG.

According to SFG, semantics mediate between the context and the lexico-grammar (Teich 1999). Of particular interest to this paper is the second stratum, which is the lexico-grammar. One of the main propositions of SFG is that it considers lexis and grammar as the two ends of a single continuum, rather than two different entities. The only difference between vocabulary and grammar according to SFG is that the former expresses specific meanings and the latter expresses more general meanings (Halliday and Hasan 1976; Halliday and Matthiessen 2014). A lexicogrammar stratum can be presented in the form of a cline similar to the one in Figure 3.

Figure 3.

Lexicogrammar cline.

As far as cohesive devices are concerned, they are distributed across the lexico-grammar cline, where reference, ellipsis, and substitution are grammatical; reiteration and collocation are lexical; and conjunctions somewhere between the two. These devices, which are part of the lexico-grammar stratum, play major roles in texts’ unity and organization. However, when considering the organization of language itself, the principle of instantiation has to be explained.

4.4. The principle of instantiation

According to SFG, any text is an instance of some underlying system. If someone does not know the system of the Arabic language, for example, a text written in that language may not have meaning to him/her. Halliday and Matthiessen (2014) compare system and text to climate and weather, respectively. Text is similar to weather in that it goes around us all the time affecting our daily lives, whereas system is analogous with climate since system underlies the impact of text. Halliday and Matthiessen (2014) state that “the relationship between system and text is a cline – the cline of instantiation” (27). The authors explain that while system represents the overall potential of language, text is the particular instance. Between the two there are intermediate patterns. A single text, or an instance of system, can be initially studied and then other texts that share certain criteria with it examined, describing this within text type. Looking at text type is seen as a movement along the instantiation cline from the instance pole to the system pole (Halliday and Matthiessen 2014). While the principle of instantiation is concerned with organization of language, the fifth principle of SFG is linked to the three metafunctions of language.

4.5. The principle of metafunction

SFG asserts that one primary function of language is to make sense of experiences, therefore construing human experience. Hence, language names and categorizes things. Language also develops categories into further taxonomies. For example, building is a category that includes houses, towers, schools, cottages, etc. Animal is another category that includes camels, lions, and so on. In Arabic, taxonomies can be categories of their own as well because they can be broken down into further taxonomies. For example, جمل/jamal/, meaning camel, which is a taxonomy of حيوان/Hayawaan/, meaning animal, can become a category in its own right as there are approximately a 100 sub-types for this animal in Arabic. Table 1 provides some examples of camel taxonomies in Arabic.

According to SFG, the language function that involves construing human experience is called ideational. The second function of language is the interpersonal, which involves “enacting our personal and social relationships with the other people around us” (Halliday and Matthiessen 2014, 29). This function of language entails that a clause of grammar exceeds being a representation of some process as it also entails some kind of function, such as offering, expressing opinion, and informing, to name but a few. Akin to this view of the interpersonal metafunction of language, Bonyadi (2011) and Fowler (2003) maintain that language does not allow its users to say something without conveying some kind of attitude, or point of view towards what is being said.

The above two functions, construing experiences and enacting interpersonal relationships, call for a facilitating function, hence the textual function of language. This function enables the other two to construct sequences of discourse, organize the flow of ideas, and create cohesion (Halliday and Matthiessen 2014). The textual function of language is divided into structural, or syntagmatic, and non-structural, or paradigmatic, components. Cohesive devices, the focus of the present paper, are pinned into the non-structural component of the textual function of language as illustrated by SFG. In a nutshell, Halliday and Hasan (1976) explain that the ideational component of language expresses content, whether it is experiential or logical, that the interpersonal component represents the speaker’s attitudes and judgments, and that the textual component represents the forming of the text in the linguistic system. The time is probably ripe at this stage to illustrate the conceptual framework of cohesion, which will be utilized to build the new model of cohesion.

4.6. Conceptual framework of cohesion

The conceptual framework of the studies reviewed in this paper serves two main purposes. Firstly, it explains the concepts that constitute the core elements of the comprehensive model of analysis suggested by this study. Secondly, it highlights the developments and adaptations that have taken place since the 1976 model of cohesive devices was first introduced. According to Halliday and Hasan (1976), cohesion “…refers to relations of meaning that exist within the text, and that define it as a text” (4). The following conceptual framework of cohesive devices is primarily discussed in light of Halliday and Hasan (1976) lexico-grammatical model. The authors have identified five main categories of cohesion that can be grouped under grammatical cohesion (reference, substitution, and ellipsis), lexical cohesion (reiteration and collocation) and partly grammatical, partly lexical cohesion (conjunctions). The following review discusses all these categories and all the adaptations and additions that they have undergone since 1976.

4.7. Reference

According to Halliday and Hasan (1976), reference involves the use of textual elements that cannot be decoded in their own right. The authors identify personals, demonstratives, and comparatives as examples of this category. They explain that these items fall within two broad reference types, exophoric and endophoric. The latter, according to them, can be further divided to anaphoric and cataphoric. Figure 4 is a rough representation of these categories.

Figure 4.

Types of reference (Halliday and Hasan 1976).

According to Widdowson (2004), exophoric reference looks outside the text to decode the identity of the linguistic item being referred to. Consider the following example:

The two referring items, The and there, in [5] cannot be decoded except by going outside the text to consider the specific context, or the shared world between the speaker/writer and the hearer/reader. It is immediately clear, though, that Halliday and Hasan’s model addresses exophoric reference as exclusively situational, and context specific. However, exophora quite often extends beyond the situation to encompass society and culture. Therefore, Paltridge (2012) introduces homophoric reference, “where the identity of the item can be retrieved by reference to cultural knowledge, in general, rather than the specific context of the text” (116). Following is an example from Arabic:

In [6] it is not possible to decode the referent of the Iraqi Chobi without knowledge of the Iraqi culture, particularly in this instance that the Chobi is a folkloric Iraqi dance, usually performed in weddings and particular celebrations. Since this decoding process requires knowledge of culture, rather than the specific context of the statement, reference is homophoric.

Figure 5 presents the adapted types of reference based on Halliday and Hasan (1976) and Paltridge (2012).

Figure 5.

Types of reference .

Endophoric reference, on the other hand, involves ties within the text and can be anaphoric, where the interpretation of the linguistic item involves moving back, or cataphoric, where decoding the reference calls for a forward movement in the text (Halliday and Hasan 1976). Consider the following examples:

In [7] all the boldfaced referring items are instances of anaphora since they can only be interpreted by going back in the text, whereas in [8] He is cataphoric because its interpretation involves moving forward in the text.

Cutting (2008) adds that endophora can be represented in terms of associative, co-textual relations in addition to the direct anaphoric and cataphoric representations. By way of elaboration, Cutting (2008) introduces the following example (10):

In example [9], in order to infer that video sharing, meaning public viewing online, is NOT physically passing DVDs to friends, readers have to rely on their knowledge of the “presuppositional pool of ‘website’” (Cutting 2008, 10). Associative endophora, then, entails that a noun phrase is linked to entities that are associated with another noun phrase in the same text. Since this type of endophora was not introduced in the 1976 model of cohesion, it will be added to the model that will be developed in this paper. Figure 6 incorporates this adaptation.

Figure 6.

Types of reference (Halliday and Hasan 1976; Cutting 2008; Paltridge 2012).

Halliday and Matthiessen (2014) divide reference expressions into two major groups: co-reference, where what is presupposed is the same referent, and comparative reference, where the presupposed is another referent of the same class. Personal and demonstrative pronouns are examples of co-reference, whereas comparative adjectives and adverbs are examples of comparative reference.

In Arabic, all the above categories of reference hold; nevertheless, English and Arabic are very much different in their linguistic structures and textual features (Alfadly and Aldeibani 2013), which is conspicuous in the number of personal pronouns in both languages (Wightwick and Gaafar 2005). While English, for example, has 7 subject pronouns, Arabic has 14. Table 2 illustrates the categories of subject pronouns in both languages.

Even a cursory glance at Table 2 will confirm the observation made by Wightwick and Gaafar (2005, 15) that: “Arabic has more pronouns than English since it has different versions for masculine and feminine, singular and plural, and even special dual pronouns for two people or things.” It is immediately clear from Table 2 that it has two functionally corresponding pronouns in Arabic, you five, and they three. The bigger number of Arabic personals does not mean that identifying referential ties in Arabic is more complicated than in English since in both languages the referent of the pronoun can be decoded exophorically, homophorically, or endophorically. Still, the above set of personal pronouns contains a major difference between the two languages as far as this type of tie is concerned. This difference will be delineated in the following sections.

4.8. Ellipsis

Ellipsis is a cohesive device that involves the omission of linguistic items which can be retrieved from another clause (Hoey 2001). Because of the omission feature, “ellipsis can be thought of as a ‘zero’ tie because the tie is not actually said” (Hatch 1992, 225), yet something is presupposed by means of what has been expunged (Halliday and Matthiessen 2014). The 1976 model of cohesion identified three types of ellipsis, which are nominal, verbal, and clausal, a categorization that has been broadly acknowledged by a number of authors and researchers (e.g., Jabeen, Mehmood, and Iqbal 2013; McCarthy 1991). Following are examples that represent the three categories of ellipsis:

In [10], the ellipsis is nominal since the deleted item is the noun fault. [11] is an example of a verbal ellipsis with part of the verb deleted, and finally, [12] is an instance of clausal ellipsis since the entire clause that normally follows Yes in such answers is deleted. As for Arabic, all the three types of ellipsis exist, yet with some difference vis-à-vis nominal ellipsis, which is not common in Arabic except when the subject of the sentence is dropped in certain cases, which will be discussed later when talking about Arabic as a pro-drop language. As for verbal ellipsis, [11] can be translated to Arabic maintaining the verbal ellipsis as follows: تستطيع أن تفعل ذلك. أنا متأكد أنها تستطيع/tastaTii`u ’an taf`ala dhaalik. ’anaa muta’akkidun ’annahaa tastaTii`/, deleting أن تفعل ذلك/’an taf`ala dhaalik/, which is equivalent to “do it” in [11]. The clausal ellipsis instance in [12] can also be rendered as is in Arabic as follows: هل وصل؟ أجل./hal waSala? ’ajal/, thus crossing out the same parts deleted in English.

Despite the agreement on the three broad categories of ellipsis, a number of issues have emerged regarding this cohesive device. One of those issues is whether ellipsis is always anaphoric or not. A number of researchers emphasize that ellipsis can be merely described in terms of anaphora because the omitted item(s) can only be retrieved by moving backward in the text (Halliday and Hasan 1976; Halliday and Matthiessen 2014), like the movement done in [10], [11], and [12] above. In accordance with this claim, Crystal (2006, 43) maintains that ellipsis “can be recovered only from the preceding discourse.” However, Jones (2012) and McCarthy (1991) confirm that English does have cataphoric ellipsis; McCarthy (1991) provides the following example (43):

Retrieving what has been omitted after could requires a forward movement. Accordingly, ellipsis can be used cataphorically in front-placed subordinate clauses. In Arabic, ellipsis can be described in terms of cataphora, too. The following is an example from Arabic; the English word-for-word translation is also provided.

The Arabic statement in [14] is functionally equivalent in English to it was not my decision, moving to the new house. In order to retrieve the speaker’s decision, one needs to move forward in the text, which makes this statement an example of cataphoric ellipsis.

Ellipsis was subject to further investigation when Thomas (1987) added more details to the category of verbal ellipsis by further dividing it into two types as far as form is concerned: echoing and auxiliary contrasting. While the former involves using part of the verbal phrase that is just before the omitted part, the latter involves changing the grammatical set of the auxiliary verb into another. The following are examples of echoing and auxiliary contrasting presented, respectively, in [15] and [16]:

As far as Arabic is concerned, it differs from English in that it is a pro-drop language, which means that the subject pronoun can be deleted because Arabic’s rich verbal morphology allows for this, in what is sometimes referred to as zero anaphora (Ryding 2005). Consider the following example:

It is immediately clear that he, the subject pronoun, of the verb visits is dropped from the Arabic text, which is still grammatically correct in Arabic. This means that in Arabic, this cohesive tie, which is the deleted-yet-retrievable subject pronoun, does not have to be physically present in the text. In English, a statement like “Ahmed is fond of his grandmother, so visits her a lot” is ungrammatical, whereas in Arabic it is grammatical.

4.9. Substitution

Substitution is very much similar to ellipsis except that an explicit indication is given that something has been deleted (Halliday and Matthiessen 2014). Put differently, it is a structural relationship that involves the replacement of one item by another (Jabeen, Mehmood, and Iqbal 2013). Like ellipsis, substitution falls into three categories: nominal, verbal, and clausal. The following are some examples:

In [18] one replaces the noun bag, and it is, therefore, an instance of nominal substitution. [19] contains an example of verbal substitution with do substituting for go to the party. Finally, so in [20] replaces an entire clause, You look tired or You are tired which is why it is an instance of clausal substitution. In Arabic, the three substituting items, one, do, and so, exist at the levels of form and function, with the words واحدة/waaHida/, تفعل/taf’al/and كذلك/kadhaalika/that can function the same as the three substituting items in [18], [19], and [20], respectively. Halliday and Matthiessen (2014) maintain that one and do are the most common nominal and verbal substitution items, respectively, whereas so and not are the most common for clausal substitution. However, other words can be used to substitute. For example, McCarthy (1991) provides the following example in which the same is used to substitute a noun:

The substituting item in [21] can be functionally rendered in Arabic using two words, such as الشيء نفسه/ashshay’a nafsahu/, which can be literally translated into the same thing, thus adding one more word “thing”, which was not structurally necessary in English.

There are still two issues to consider with this cohesive device. The first one is whether it is always anaphoric as claimed by Halliday and Hasan (1976) and Halliday and Matthiessen (2014). In fact, there is no reference to substitution as a cataphoric device in the literature so far. However, in certain types of texts, it seems, one can be cataphoric when it is preceded by a demonstrative this. In this case, one no longer replaces a noun but a general idea. Following is an example from a New York Times op-ed:

In this example, this one refers forward to the Italians tweets that mock ISIS’s warning of heading to Rome. The point is that in certain cases, substitution can be cataphoric. The second point is that the lexical items introduced in this section one, do, and so are not always substitutive. On this, Salkie (1995) provides the following examples (36):

This realization also has grounds in Arabic as واحد/waaHid/(Arabic for one) can have a numeric value, تفعل/taf`al/(Arabic for do) can have the mere function as the main verb in the sentence, and جِدّاً/jiddan/(Arabic for so) can function as a modifier, meaning very.

4.10. Conjunctions

This particular category of cohesive devices has undergone several adaptations since its introduction by Halliday and Hasan in 1976 (Ahangar, Taki, and Rahimi 2012). The reason for this could be that it is not easy to produce an exhaustive list of the entire range of conjunctions (McCarthy 1991). Therefore, the 1976-model of conjunctions, which consisted only of the categories of adversatives, additives, causal, and temporal, went on an adding-up spree that may never come to a decisive end. Halliday and Matthiessen (2014) expanded the four types of conjunctions into nine by adding apposition (e.g., in other words, for example), clarification (e.g., in short, by the way), variation (e.g., instead, except for that), comparative (e.g., similarly, in a different way) and respective (e.g., in this respect, elsewhere). Locke (2004) added one more category, listing, and argues that temporal conjunctions, such as first and second, can also serve listing purposes since they can be used to list the elements of an argument. These expressions, representing the category of listing, are acknowledged to have identical functions in Arabic, too (Lahlali 2009). Table 3 presents the 10 categories of conjunctions with examples from English with Arabic alongside. These categories and examples are adapted from Halliday and Matthiessen (2014), Haywood and Nahmad (1993), Lahlali (2009) and Locke (2004).

The above sets should not lead to the conclusion that English and Arabic have entirely identical sets of cohesive devices because each language has its own particular system. For example, the one-letter conjunctions وَ/wa/ and فـَ/fa/may have a variety of English correspondences belonging to different sets based on the context they are used in (Abu-Chacra 2007; Haywood and Nahmad 1993).

4.11. Lexical cohesion

According to Smith (2003), lexical cohesion is a seminal contribution of Halliday and Hasan (1976) as it has enabled linguists to find patterns of lexical co-occurrence in texts. However, the two categories of lexical cohesive devices, reiteration and collocation, which appeared in Halliday and Hasan’s 1976 model, witnessed several adjustments, which have been integrated into the model suggested in this paper. Basically, all the developments and adjustments to the 1976 model maintained the category of reiteration, which involves repetition of the same word, while some of them have raised questions about collocation, describing examples of it as arbitrary co-occurrences, thus excluding them from lexical cohesion analysis (Hasan 1984; McCarthy 1988), or including them with certain adjustments that seek to systematize them, yet acknowledging the difficulty of doing so (Tanskanen 2006). It has been agreed, though, that collocation generally refers to the association that links the words that co-occur, or that have the tendency to occur with each other (Sinclair 1991; Stubbs 2001). One development at the level of reiteration is that introduced by Hoey (1991), who divided repetition to two categories, namely simple lexical repetition, such as a girl/girls and complex lexical repetition, referred to by de Beaugrande and Dressler (1981) as partial recurrence, such as drug/drugging. Scott and Tribble (2006) view repetition as a cohesive device in terms of keyness, whereby lexical items that reflect what the text is about are reiterated to signal their importance. One of the most comprehensive models of lexical cohesion has been developed by Halliday and Matthiessen (2014), who have divided lexical cohesion into five categories, which are repetition, synonymy/antonymy, hyponymy, meronymy, and collocation. It might be claimed that this 2014 model is built on a previous categorization introduced by Martin (1992), a classification that encompassed all the lexical types included in the 2014 model except for collocation. Therefore, the general classification of the 1976-model still holds yet with the addition of synonymy/antonymy, meronymy, and hyponymy as categories within it. Within the category of synonymy/antonymy, Halliday and Matthiessen (2014) have maintained the subcategory of general nouns, firstly introduced by Halliday and Hasan (1976), thus introducing lexical items, such as thing, stuff, and place, which can be synonymous with other lexical items in the text in certain situations. This particular group has been referred to in the literature using various terminology, such as signaling nouns (Flowerdew 2003) and shell nouns (Aktas and Cortes 2008). It should be noted here that the subset of synonymy/antonymy has been emphasized as a major cohesive type in Arabic as it is usually used to express a wide range of meanings (Parkinson 2006). Table 4 shows some examples of lexical cohesive devices in English and Arabic.

4.12. Parallelism

This cohesive device was not introduced either in Halliday and Hasan (1976) or in Halliday and Matthiessen (2014) models. Yet, parallelism, which by and large refers to the repetition of a certain form or structure for the purposes of emphasis and insistence (de Beaugrande and Dressler 1981), has been acknowledged as a cohesive device by many scholars and authors (e.g., Neumann 2014). As for Arabic, Dikkins, Hervey, and Higgins (2002) assert that parallelism as a cohesive device typically involves repetition of the same grammatical category or categories, and that it is not as common in English as it is in Arabic. If this claim is true, it can explain the absence of this device from the 1976 model and from several other subsequent models of cohesion in English. Adding emphasis to this point, the authors suggest that Arabic-to-English translators should therefore be advised to use summary phrases instead of retaining all the elements of the source Arabic parallels when rendering parallel structures from Arabic to English. This study has added parallelism to the model suggested in this paper because it is acknowledged as a cohesive device not only by English scholars (e.g., de Beaugrande and Dressler 1981) but also by Arabic researchers (e.g., Aziz 2012).

In his analysis of the poetry of the famous Palestinian poet Mahmoud Darweesh, Sultan (2011) cites a number of examples where the poet employs parallel structures as cohesive devices. One of these examples can be found in [24] below, where the form “imperative + adverbial phrase” is repeated in a parallel structure.

Following from the above discussion on the developments of the categories of cohesive devices since they were introduced in 1976, it becomes obvious that in order for a model to be comprehensive it should consider all these changes. This does not mean that all studies have to use such a comprehensive model in their analysis of cohesive devices as whether to use it in its entirety depends on the research questions and purposes.