1There are many types of linguistic theory but, I feel, they all fall into three broad groups. Some linguistic approaches are formal in nature, and the obvious example of this is generative grammar and its various developments and offshoots, initiated by the work of Chomsky (1957, 1965). There are cognitive approaches, which would include the American West Coast cognitive school (Langacker 1982, 1987), as well as the French Théories de l’énonciation (Culioli 1990, 1999a, 1999b). Finally, there are functional theories, of which there are several varieties (Butler 2003a, 2003b; François 2018), such as Role and Reference Grammar (van Valin 2001), the Functional Grammar of Dik (1985), and, in France, Martinet’s Functional Linguistics (1985). But the functional approach which I have found most conducive to my own purposes is that initiated by Halliday in the 1960s and then known as Scale and Category Grammar, but which developed into what is now known as systemic functional linguistics (Halliday 2014; Thompson 2004; Banks 2019).
2Cognitive and functional approaches are in many ways compatible. According to Halliday and Matthiessen, while cognitive approaches go from cognition to language, their functional approach goes from language to cognition: “Instead of explaining language with reference to cognitive processes, we explain cognition by reference to linguistic processes” (Halliday & Matthiessen 1999: x). So, cognitive and functional approaches are, as it were on the same road, but while one is travelling from A towards B, the other is travelling from B towards A, so they must meet somewhere. Formal approaches are somewhat apart, particularly since, in the wake of Chomsky, they tend to have as their object of study an idealized, and hence artificial, language, rather than a natural language. At the same time, all languages are encoded in form, so all theories will have to deal with form at some point, so there may be a meeting point even in this case.
- 1 An earlier version of this paper was given at the Study Day Anglais de spécialité et linguistique, (...)
3My object in this paper1 is to attempt to show some of the ways in which I have found systemic functional linguistics to be useful in the study of specialized text. This is not intended to imply that this is to the exclusion of other approaches. All of the features that I shall look at can be studied within other frameworks, but I have found that systemic functional linguistics lends itself to this type of study. I shall take as my example for illustrative purposes the case of the physical sciences research article.
4Until fairly recently it was commonly believed that in the scientific research article writers avoided the use of first-person pronoun subjects in favour of the passive form. A number of studies (Barber 1962; Huddleston 1971; Banks 1994) suggested a passive rate of the order of 30% of finite verbs. However, over the last 15 years or so some studies have suggested that there is evidence of a fall in passive use (Seoane 2006; Seoane & Hundt 2018), while others suggest a rise in the use of first-person pronoun subjects (Hyland & Jiang 2016a, 2016b, 2017). In an attempt to see to what extent these claims are valid, I recently (Banks 2021) took a small sample of six randomly chosen articles from the Proceedings of the Royal Society A for the year 2018. This journal deals with the mathematical, physical and engineering sciences. The articles have between two and six co-authors. None of them are single-authored. The total number of co-authors is 23, 14 of whom have typically anglophone names. There are 13 institutional addresses provided by these authors, all but one in the UK, the exception being an address in Switzerland. For each article the journal gives three subject areas. The area of civil engineering occurs three times and environmental engineering and mathematical modelling twice each. The others were acoustics, applied mathematics, artificial intelligence, electrical engineering, energy, fluid mechanics, geophysics, hydrology, mathematical physics, structural engineering, and wave motion.
5The sample has a total of 2808 finite verbs, of which 825 are passive in form, giving an overall passive rate of 29%, which corresponds to the rate that had been prevalent during the 20th century. Taken on its own, this would seem to indicate that the rate of passive use has not changed. However, the overall rate masks the fact that there were wide differences between individual articles, where the passive rate varied from 11% to 54%. Two of the articles had relatively low rates (11%, 17%), two were comparatively high (41%, 54%), while two hovered somewhere in the middle (both 27%).
6Consequently, I decided to consider the finite passive forms in terms of the process types that they encoded. I use a system of five process types derived from those used in systemic functional linguistics (Halliday 2014, Banks 2019). Material processes encode physical actions and events.
(1) The air phase is not directly driven by the capillary pressure, but can be set in motion by the water velocity at the air-water boundary. (Cooper et al. Proceedings A 2018)1
7Mental processes encode events of a cerebral nature.
(2) In this study, the limiting material stress has thus far been assumed to be the same in tension and compression. (Fairclough et al. Proceedings A 2018)
8Relational process link two entities, or an entity with one of its characteristics.
(3) However, such a formulation is restrictive as it is not valid for magnetic media in the electromagnetism context or for scatterers with varying density in acoustics, as identified in [19]. (Gower et al. Proceedings A 2018)
9Verbal processes are processes of communication.
(4) For instance, Zheng et al. [27,28] proposed multi-stage approaches involving the decomposition of the water network using graph algorithms and then the optimization of the decomposed networks using linear and nonlinear programming and DE. (Zhao et al. Proceedings A 2018)
10Existential processes are statements of existence.
(5) That is, in the limit where there are no cylinders, except one fixed at x2, the averaged scattering coefficient An tends to the scattering coefficient of one lone cylinder, even for 0 < x2 < ẋ. (Gower et al. Proceedings A 2018)
11A question arises as to the nature of the processes expressed by “can ... be calculated” in (6) and “have combined” in (7).
(6) The resilience to uprooting and the prediction entropy can also be calculated starting from the rooting depth and the expected scouring dynamics. (Perona et al. Proceedings A 2018)
(7) This is an advancement from the work of Daly & Roose [8] as we have combined the equations for fluid flow with the equations for exudates diffusion. (Cooper et al. Proceedings A 2018)
12The verbs in question here deal with mathematical calculation, and, I would argue that this can be encompassed within the notion of mental process. Mental processes have often been subdivided into cognitive, perceptive, and affective types, to which some would add a desiderative type (Thompson 2004). It could be argued that mathematical processes are a type of cognition process, but the prevalence of mathematical processes in this genre means that it is useful to isolate mathematical processes as a distinct sub-category. Moreover, I would extend this category to include cases where a computer is used. Whether a mathematical calculation is carried out in the head, using pen and paper, or a more sophisticated machine like a computer, does not alter the essential nature of the process. I would therefore include an example like “was ... simulated” in (8) and “have been conducted” and “are solved” in (9) in my count of mental mathematical processes.
(8) After the above initial trials and given the proposed methodology, the optimization of the water network operation was then simulated from 30 April 2016 to 30 January 2017. (Zhao et al. Proceedings A 2018)
(9) The simulations have been conducted using a well-known DNS code SENGA [11-17], where the governing equations of mass, momentum, energy and reaction progress variable are solved in non-dimensional form. (Wacks et al. Proceedings A 2018)
13Processes can also be encoded by non-finite verb forms and by nominalized processes. The question of nominalizations will be considered below. Here I am dealing with passive forms, and later verbs with first person pronoun subjects. Both of these require a subject, and while non-finite forms may occasionally have a subject or a putative subject that can be recovered from the co-text, this is not typical, and I shall not deal with non-finite verbs forms here. A process type analysis of the finite passive verbs gives the following result (see table 1).
Table 1 Process types of finite passive verbs
Type
|
N
|
%
|
Mental
|
508
|
62
|
Material
|
171
|
21
|
Verbal
|
115
|
14
|
Relational
|
41
|
4
|
14In addition, the mathematical subcategory alone accounts for 37% of all finite passive verbs. Hence, in this sample, where passive verbs are used, almost two-thirds of them are examples of mental process, with over half of these being of the mathematical subtype. Passive material processes are much rarer, accounting for about one passive verb in five.
15I would now like to turn to the related question of the use of first person pronoun subjects. The articles in this small sample have between three and six co-authors. There were no single-authored papers, and it might be pointed out that single-authored papers have become relatively rare in the physical sciences. Consequently, the first person pronoun subjects in question here are all examples of the plural form we. Of the 2808 finite verbs in this sample, 301 have first person pronoun subjects, that is, 11% of the finite verbs. This is much less than the rate for passive forms, but still quite large if one takes something approaching zero to be the norm. However, as was the case for passives, the overall figure masks wide discrepancies in individual articles. In one of the articles, the rate for first person pronoun use was a relatively high 33%; in two others the rates were more modest, 17% and 12%. So, these three could be said to use first person pronoun subjects, one of them extensively. On the other hand, the other three could be said not to use them. One had two examples for 483 finite verbs; another, a single example for 552 finite verbs; and the last had none at all.
16This suggests that there may now be two (or perhaps more) models available for the writing of the physical sciences research article: one with extensive passive use and few first person pronoun subjects, and one with fairly free use of first person pronoun subjects, and less passives. It probably also indicates that the situation is in a state of flux at the moment.
17If it is the case that first person pronoun subjects can be used in the research article, it is not sufficient to know that they are used: we need to know when they are used. An analysis in terms of process types can cast light on this question. The analysis of the process types of verbs with first person pronoun subjects in this sample gives the following results (see table 2).
Table 2 Process types of verbs with first person pronoun subjects
Type
|
N
|
%
|
Mental
|
228
|
76
|
Verbal
|
56
|
19
|
Relational
|
13
|
4
|
Material
|
4
|
1
|
18Here we see that of the first person pronouns which occur, three-quarters do so with mental process verbs. Moreover, the majority of these are of the mathematical type, which account for half (50%) of the verbs with first person pronoun subjects. Verbal process accounts for a much more modest 19% of the verbs with a first person pronoun subject. First person pronoun subjects occur rarely with relational processes (4%), and with material processes hardly at all (1%). The fact that first person pronoun subjects hardly ever occur with material processes is striking, and, I think, significant.
19It might also be noted that mental and verbal processes taken together account for 95% of the verbs with first person pronoun subjects. Mental and verbal processes have much in common in that they are particularly human occurrences. Indeed, in the early days of the development of systemic functional linguistics, what we now call verbal process was considered a subtype of mental process, then called “mental externalized process” (Berry 1975). Even today, the variant of systemic functional linguistics known as Cardiff Grammar does not distinguish between mental and verbal process (Fawcett 2000a, 2000b; Bartley 2018).
20It therefore seems possible to say that one of the possible models for writing the physical sciences research model would include the use of first person pronoun subjects, but mainly with mental process verbs, particularly of the mathematical type, and to a lesser extent verbal process. For practical purposes, it might also be said that first person pronoun subjects do not occur with material process verbs.
21One might well ask why this should be the case. I suspect that the reasons lie in the history of the development of the scientific research article (Banks 2008). A possible explanation might run along the following lines. Until the mid-nineteenth century the physical sciences research article was mainly experimental, with extensive use of material processes, as in this example from 1820.
(10) When the powder is boiled in alcohol, it is partially decomposed, and assumes a lighter colour; if it be then thrown on a filter, the odour of acetic acid is soon perceived … (Davy, Transactions 1820)
22From the mid-19th century onwards an increasing use of mathematics began to change the nature of the physical sciences research article, as in this example from 1880.
(11) For prisms I, II, and IV, the value of ϑ is given by formulæ of the form Θ=λ±ϕ’,Φ’ having the meaning attached to it in the results of experiment. (Glazebrook, Transactions 1880)
23This phenomenon gathered speed during the 20th century, as in this example from 1940.
(12) The energy values in columns (a) and (b) have been calculated from the wave functions obtained by the solution of Fock’s equations with exchange but without the superposition of configurations, which has also been omitted in the energy formula in obtaining the results in column (a), but included in the energy formula only, in obtaining the results in column (b). (Hartree et al. Transactions A 1940)
24And this was compounded from the mid-20th century onwards by the introduction of computers and mathematical modelling, as in this example from 1985.
(12) In the theoretical part of this article (§§2-5), a model is developed for wind-generated topographic waves in an elongated two-layer lake in which the upper layer thickness D1 is small by comparison with D2, the characteristic depth of the lower layer. (Mysak et al. Transactions A 1985)
25Indeed, it is frequently difficult for the non-specialist to determine whether what is being talked about is the phenomenon, or the model of the phenomenon, and one may be led to ask whether the scientist himself makes this distinction. Perhaps for the scientist, the phenomenon and the model are virtually the same thing, or, at least, the difference is not pertinent for him.
26This development is probably also aided by the fact that mental processes are specifically human, and indeed specific to a particular individual. Someone else may repeat my physical actions, but no-one else can have my thoughts!
27Some readers might wonder whether it would have been useful to note the sections in which the various phenomena studied are to be found. However, the sections of these articles do not fall easily into the well-known IMRD structure. They all have something which corresponds to an introduction and a conclusion, but in between there is great variability, which would make comparison difficult. The section headings as they appear in the articles is shown in table 3. Articles are identified by the first named author.
Table 3 Section headings
Cooper
|
Fairclough
|
Gower
|
Perona
|
Wacks
|
Zhao
|
1 Introduction
|
1 Introduction
|
1 Summary
|
1 Introduction
|
1 Introduction
|
1 Introduction
|
2 Derivation of equations
|
2 Elements of equal strength
|
2 Multiple methods for cylinders
|
2 Mathematical model
|
2 Mathematical background and numerical implementation
|
2 Preliminaries
|
3 Validation of homogenization procedure
|
3 Layout optimization
|
3 Averaged multiple scattering
|
3 Model validation
|
3 Results and discusssion
|
3 Differential evolution
|
|
4 Application to very long-span bridges
|
4 Effective wave number
|
|
|
4 The methodology for water network operational optimization
|
|
|
5 Two species of cylinders
|
|
|
5 Experiments
|
|
|
6. Numerical examples
|
|
|
|
|
|
7 The average field and reflection
|
|
|
|
4 Conclusion
|
5 Conclusion
|
8 Conclusion
|
4 Conclusion
|
4 Conclusion
|
6 Conclusion
|
28It may be that the difference arises from the fact that articles in the field are now mainly a question of mathematical modelling, whereas the IMRD structure is probably more appropriate for experimental articles. Moreover, many of the headings here appear to relate to questions specific to the individual article. Consideration of the extent to which the structures of these articles can be assimilated to the IMRD structure, or whether a different type of structure needs to be elaborated for this type of article would provide a research project in itself.
29I would now like to turn my attention to another frequently mentioned feature of scientific writing, that of nominalization of processes. This was already fairly common in the eighteenth century when examples occurred at a rate of about one per 30 words of running text.
(14) … and the interval QP will be the greatest Aberration or Error, occasioned by the Separation, or unequal Refraction, of the greatest and least refrangible Rays, after their Emergence from the concave surface FED. (Smith, Transactions 1740)
30This increased rapidly during the 20th century to attain a rate of about one per 15 words of running text.
(15) When heat is added to the flow in such a channel, the Mach number becomes closer to unity whether the flow is supersonic or subsonic. When the Mach number in unity, no further addition of heat is possible, without causing a breakdown of the flow. (Axford, Transactions A 1960)
31From the mid-20th century onwards, nominalized processes began to be used fairly extensively with a modifying (adjectival) function.
(16) The dashed lines, C, are the loci of possible transitions, via an ionization front, from various points on the strong shock curve. (Axford, Transactions A 1960)
32In (16), the nominalized process “ionization” functions as a modifier in the nominal group “an ionization front”, and the nominalized process “shock” functions as a modifier in the nominal group “the strong shock wave”.
33In systemic functional linguistics nominalization is considered to be a form of grammatical metaphor. A grammatical metaphor occurs when something is encoded in a non-congruent form. For example, the congruent, normal, way of encoding a process is by using a verb. If some other form is used, e.g., a noun, then this is non-congruent and constitutes a grammatical metaphor (Halliday 2014; Ravelli 1988, 2003). This has grammatical consequences in that the nominalized form can be used in any of the normal nominal functions, subject, complement, prepositional completive, and it has semantic consequences, in that it is not necessary to make the subject explicit, nor, if the verb is transitive, the complement. Moreover, since nominals normally encode entities, the nominalized form suggests something that is fixed, objective and permanent. In Halliday’s often-quoted phrase, “You can argue with a clause but you can’t argue with a nominal group” (Halliday & Martin 1993: 39).
34In stark contrast with what was found for passive verbs and for verbs with first person pronoun subjects, roughly three-quarters of the nominalized processes are cases of material process, and this seems to have been the case throughout the history of the physical sciences research article.
35Halliday points out that nominalization allows categorization and the construction of taxonomies and leads to the creation of technical terms. It also permits the compacting and condensation of complex notions into a simple unified form (Halliday 1988, 1998).
36The above may give the impression that grammatical metaphor is no more than a question of nominalization. Consequently, I would here like to insert a few remarks on a different type of grammatical metaphor which is relevant for the scientific research article. These remarks concern the use of extraposition. Almost all extraposition matrices (the “It is x...” part of the extraposition structure) are interpersonal, and many constitute grammatical metaphors of modality.
(17) It is possible to repeat this calculation to account for the effect of a finite-sized indenter, rin. (Box et al. Proceedings A 2017)
37It could be argued that the congruent form of (17) would be “We could repeat ...”, where use of the modal auxiliary is considered the congruent way of expressing modality. The extraposed form using a modal adjective then constitutes a non-congruent form, and is thus a grammatical metaphor. It might also be noted that this makes the expression of modality part of the thematic material as an interpersonal theme.
38The question of nominalization dovetails with that of thematic structure. By “thematic structure” I mean the structure of a clause in terms of a theme and a rheme, where theme is defined as the speaker’s starting point, and is realized in a language like English by the first main component of the clause (subject, circumstantial adjunct, complement or predicator). In (18), the theme of the first clause is “Our approach” and that of the second “The one assumption that is employed”. It will be noted that the head of each of these nominal groups is a nominalized process.
(18) Our approach does not rely on an extinction theorem or the manipulations of divergent integrals or series. The one assumption that is employed is the quasi-crystalline approximation. (Gower et al. Proceedings A 2018)
39The relationship between themes and rhemes throughout a text constitutes its information flow, and this forms part of the argument structure of the text. It is possible for the writer to introduce a new theme, unrelated to the previous discourse, but otherwise the theme can be derived from a previous theme or a previous rheme. Where the theme is derived from the theme of a previous clause, this constitutes constant progression, as in (19).
(19) The equations derived by Daly & Roose [8] are appropriate for modelling bulk soil; However, they might not be directly applicable in the region of soil close to the roots over which the plants have influence known as the rhizosphere [12]. (Cooper et al. Proceedings A 2018).
40Where the theme is derived from the rheme of a previous clause, this constitutes linear progression, as in (20).
(20) If the origin is placed at the highest point of the curve C1 = C2 = 0. A plot of the curve is shown in figure 1. (Fairclough et al. Proceedings A 2018).
41Nominalizations frequently function as themes and act as compressed forms of previous information. In (18), the nominalised form “assumption” condenses the idea of relying on something, which is found in the previous rheme and would be the congruent or “unpacked” (Ventola 1996) form of “assumption”.
42Constant progression is typical of description and simple narrative, and in the scientific research article would tend to occur mainly in the introduction and literature review.
(21) Furthermore, temperature inhomogeneity is observed in the burned gas for non-unity Lewis number flames because of the inequality of the diffusion rates of species and heat ..., whereas the burned gas temperature remains equal to the adiabatic flame temperature for the unity Lewis number flames. The effect of this temperature inhomogenity is relatively more prevalent for the Le < 1.0 flames than in the Le > 1.0 cases because the higher thermal diffusion rate in the Le > 1.0 flames tends to nullify thermal inhomogeneities in the burned gas. The temperature inhomogeneity in the burned gas gives rise to a considerable dilatation rate within the burned gas beyond the flame and this tendency strengthens with decreasing Lewis number. (Wacks et al. Proceedings A 2018).
43Linear progression tends to occur in texts which are building up an argument.
(22) Homogenization is based on a linear expansion of the dimensionless equations in terms of ε. This leads to a cascade of numerical problems, in which the details of the pore scale geometry are captured by solving representative cell problems on one period of the domain [25]. The results from the cell problems then parametrize averaged macroscale equations that approximate the solution of the full set of equations. Importantly the homogenized equations are solved on an averaged geometry ... (Cooper et al. Proceedings A 2018)
44All texts present a mixture of constant and linear progression, and long sequences of one or the other are rare, but in general, linear progression tends to be more frequent than constant progression in the physical sciences research article.
45So far, I have discussed the incidence of passive forms and first person pronoun subjects in terms of process types; I have also considered nominalized processes and its relationship with thematic structure. I have done this within the framework of systemic functional linguistics, but this does not mean that similar studies could not have been carried out in terms of other theoretical approaches. I have used the concept of process types, but many approaches use various typologies of verbs, although it should be noted that process is a function, not a form. Nominalizations have also been studied from a number of other points of view. The concept of thematic structure has been used in a number of schools of thought, most notably, but not exclusively, the Prague school (Firbas 1992).
46I have found systemic functional linguistics conducive to my purposes, because I feel it gives an overall theory of language where all these features have an interrelated place, and are not simply a number of isolated phenomena. In particular, systemic functional linguistics sees the clause in terms of three metafunctions: ideational, the relationship of processes, entities and circumstances which is being expressed in the clause; interpersonal, the relationship between the speaker and his addressees, and between the speaker and his message; and textual, the way the clause is structured. Process types form part of the transitivity (in the systemic sense) of the clause, and as such fit into the ideational metafunction. Thematic structure is part of the organization of the clause and so is part of the textual metafunction. Grammatical metaphor shows how the speaker is manipulating his clause and is thus part of the interpersonal metafunction. Thus, systemic functional linguistics, far from treating these features as isolated phenomena, fits them into a total theory of language.
47I have illustrated this article using the example of the physical sciences research article. Even within the scientific field, there may be differences between various sub-genres (Biber & Gray 2013). In particular, throughout the history of the scientific research article it would seem that the development of the biological research article lags behind that of the physical sciences. First results (Banks 2017) would seem to indicate that developments in the use of the passive form and first person pronoun subjects observed in the physical sciences have not yet affected the biological sciences to the same extent.