Plain language practices of professional writers in Quebec

This article investigates the plain language practices of professional writers in Quebec, using a survey. We contacted 55 professional writers and asked them to complete an online survey about how they apply plain language in their work, and the type of writing assistance they would find useful. We also asked 40 of those writers to carry out a simplification task to see what kind of simplifications they were actually making. If the feelings about the reality of the writers’ work is in line with the literature, opinions on plain language guidelines are not. Most writers in our survey find them useful and precise enough, and this contrasts with reported criticisms of such guides. In the simplification task, we noticed that writers focus on the overall understanding of the text, and not only on some linguistic characteristics (as shown in plain language guidelines). The more experienced the writer is, the more changes they will make to visual/structural aspects or relational efficiency. Putting the focus on the reader’s needs is their main concern.


Introduction
Plain language is about making a text more comprehensible for its target audience. Plain language was initiated with the aim of making specialized texts -which are often really hard to apprehend (Schriver, 2017) -more accessible, and was first considered in the field of law. Plain language focuses on the writing process to get a simplified text (a writer works by iterations) and the characteristics of the text that make it hard to read. Researchers study methods used to simplify texts and evaluate their complexity (Schriver, 1989). These range from peer-judgment (Crépin, 2011) to tests on readers (Schriver, 2012;Crépin, 2011), readability measures (Collin-Thompson, 2014;François, 2011) and plain language guidelines based on the different representations of the reader. Those guidelines give advice to help the writer to write as clearly as possible for the target audience.
Plain language is part of the vast field of "writing studies" (in French: "rédactologie"): research that looks at professional communication, and more precisely on the processes and knowledge involved in the production of such texts, as well as their suitability for the audience (Labasse, 2001).
In this domain, we can distinguish several axes of research: focus on the writers or the readers (their vision of the world, their level of competence, etc.), focus on the writing or reading process, and focus on the analysis of the texts produced (Clerc, 2005). Plain language is part of the last axis, because its main focus is on the text itself, and the simplification of its linguistic characteristics.
A clear, simplified text has more chance to be efficient, to convey the intent of the sender to their readers (Cardinal, 2008). As the texts studied in the context of plain language research are written in a professional context, they have a practical aim: to help people understand ideas, take actions, make procedures (Schriver, 2012).

Discourse and Writing/Rédactologie
Volume 31, 2021 http://journals.sfu.ca/dwr 51 A professional writer's task is to convey the message to the addressee. As we will see in the next section, this mission is complex. A writer can rely on helpers (such as colleagues, guidelines), but is also subject to several constraints, which might be internal (lack of training, expertise, etc.) or external (external criticism of their work, lack of time, lack of support, guidelines that do not apply to the situation, etc.) (Nord, 2018).
For the past few years, writers working in administrations can also utilize new technologies.
Processing tools such as LARA (Cosla, 2001), AMesure (François et al., 2020) or PAR4SEM (Yimam & Biemann, 2018) are now available, but they are not yet based on the real needs of the writers.
Planning to create a help tool based on real writers' practices, we examined the work of professional writers. Given the importance of the task of writing and simplifying texts in their everyday work, their habits should correspond to the good practices of clear writing. We want to describe these habits, as a first step to automatize them. In interviewing the writers, we pursue three objectives: • to discover their perception of plain language: what does plain language mean to them? Do they agree with the criticisms and the limits found in the literature? Is plain language sufficiently used in administrations?
• to identify their expectations about plain writing assistance: are the existing tools sufficient?
Are plain language guidelines really useful? What form should a tool for plain language take in the future?
• to verify agreement between professional writers' habits and the advice provided in plain language guidelines: are simplifications made by professional writers the same as those suggested by the guidelines?
To answer these questions, we conducted an online survey of 55 professional writers from Quebec. We chose Quebec because it was a pioneer in professional writing in the French-speaking area . Therefore, professional writing habits have more impact than in other countries, where this type of writing is just starting to be taught, or is still absent from university curricula. We asked the writers about their plain language practices and their expectations from assistance, and we carried out a simplification task of excerpts with 40 of those writers.
As we will see in this article, professional writers tend to simplify a text in a global way, and not only focusing on specific linguistic phenomena (as suggested by plain language guidelines). The more experienced a writer is, the more he will make global changes. The vision of plain language our writers endorse to is in line with the advances in studies about communicational efficiency, especially with regard to relational efficiency (Romain et al., 2016). Discourse and Writing/Rédactologie Volume 31, 2021 http://journals.sfu.ca/dwr 52 In this article, we will first present the previous studies that led us to focus on the writers' practices. Then we will explain our method, the way our survey was conducted. The fourth section will aim to describe the results of our survey with qualitative and quantitative analysis. Then, we will end this article with a presentation of lines for future research.

Review of the literature
Research on writers' practices covers different dimensions of their work. First, as our study focuses on them, we gather studies on the characteristics of a professional writer. We will then describe what we know about their practices, especially regarding the aids and constraints for putting plain language into practice, since our study aims to relate to these topics.

Characteristics of the professional writer
Before getting to the practices of the writers of professional texts, it will be useful to consider what we mean by 'writers'. We can divide writers into two categories: professional writers and functional writers (Beaudet et al., 2016;Schriver, 2012). The differences between the two categories are, firstly, that professional writers have taken writing courses at a university and the others have not; and secondly, the time spent writing in their everyday work: redaction is the main job for professional writers, but is just an additional task for the functional ones. For example, a lawyer is not primarily a writer, but often has to write to clients to explain on-going legal proceedings. In other words, a lawyer produces professional communication as part of their work; but for the professional writer, this is a full-time job. In this work, we focus on professional writers and their practices. Let's note that professional writing training does not exist in every French-speaking region. Only Quebec and some universities in France offer one .
Professionalism can, however, be acquired by other means: self-training, internships, daily professional writing practice with the help of colleagues or guidelines, etc. (Gambier, 2016).
Nevertheless self-training presents some risks, such as applying unsuitable rules (Ganier, 2016) or developing bad habits that are hard to change with new training (Desbiens, 2008).
Professional writers have to write about different subjects for a client, which can be an institution, an organization, a company, a person, etc. The professional writer is the "shadow man" (Beaudet & Rey, 2014), involved in the writing process with the client -a plural enunciation (Fraenkel, 2001). Ideas do not originate with the writer, whose job is simply to give them shape. The writer is a kind of mediator, seeking common ground between the client's universe of constraints Discourse and Writing/Rédactologie Volume 31, 2021 http://journals.sfu.ca/dwr 53 and the reader's universe of constraints Romain et al., 2016). Even before writing, a writer has to define the needs of the client and clarify their communication process (Beaudet & Rey, 2014). He sometimes has to modify the writing task, because the client does not always have the end reader's needs in mind (Collette, 2008). The writing process always starts with the analysis of the assignment. Then, the professional writer seeks information and structures it, in order to write the text; the process ends with the revision of the text (Clerc, 2000).
The expertise of the professional writer lies in their ability to link two goals: informational efficiency and relational efficiency (Romain et al., 2016). The first corresponds to the clarity of the text (clear writing in its first meaning), the use of the complexity of the language to make the information understandable for the reader. The second combines an understanding of the communicative situation -analysis of all the components of the communication and the needs of the client and the reader -and the relationship between the client who holds the information and the reader who is seeking this information (Huet, 2013), to optimize the relationship between the client and the reader, and make the addressee of the message want to perform what they are asked to do. This "co-management" of the communication seeks cooperation and not conflict in an asymmetrical relationship (Beaudet & Rey, 2014), trying to spare the faces of the two parts through linguistic politeness. These two types of efficiency correspond to communicational efficiency ) -or interactional efficiency (Romain et al., 2016).

Studies on writers' practices
Our research focuses on the writers' practices, their vision of plain language and the type of assistance they expect. We have drawn up an inventory of research carried out on these different subjects. The main points of these studies concern the writing process, the writer's environment, how people outside the field view plain language, and the assistance provided to the writers.
Studies on the writers' practices are often interested in the writing process. Understanding the readers' characteristics while the writer does not have contact with them (Ganier, 2016), without any feedback from client or readers (Cardinal, 2008), is not easy. The recipient is often unknown or hard to identify (Fernbach, 2003). Especially because, in most cases, writers are addressing multiple audiences, with different expectations and skills, but with only one text to produce (Schriver, 2012).
Moreover, the writer is often seen as responsible for the content, not just the form of the message (Beaudet & Rey, 2014). Not being specialized in the domain for which they write, they may have a poor understanding of the information to be transmitted (Beaudet et al., 2012).

Discourse and Writing/Rédactologie
Volume 31, 2021 http://journals.sfu.ca/dwr 54 Studies on the writers' environment show that their working conditions are not always ideal.
The client is not always aware of the time required to complete the assignment (Beaudet & Rey, 2014). The writer may face pressure because of this, and fail to write in a manner that meets the required level (Nord, 2018).
People outside the writing profession often have a negative view of plain language, even if they write a lot in their job. Kimble (1994Kimble ( , 2016 and Pease (2012), among others, have collected many criticisms of plain language mostly from lawyers. They talk about impoverishment of the language, about a focus on particular words or on sentence length, about lack of precision in a specialized domain, about the problems of interpretation which result from it, about time-consuming activity without a real impact on understanding, and so forth. Kimble (1994Kimble ( , 2016 and Pease (2012) systematically refute these criticisms by proving the added value of plain language. However, these criticisms are persistent and the writers must also fight against these preconceptions on a daily basis.
Research on the assistance provided to writers for writing and revision of texts focus particularly on plain language guidelines. These guidelines are used in certain institutions (European Union, 2011) and governments (Ministère de la Communauté Française de Belgique, 2010; Gouvernement du Québec, 2006;etc.), and are intended mainly for writers who have not had training in plain language but whose writing work is important. Guidelines are often created with the view of a "typical" reader profile, and rely upon knowledge from psycholinguistics about the reading process (Schriver, 1989), from the field of readability, which describes the linguistic characteristics that make a text readable or not (see Chall & Dale, 1995or François, 2011 or, to a lesser extent, from studies about the impact of text simplification on the reading speed and comprehension (L' Allier, 1980). Guidelines focus on informational efficiency, clarity of the message, reduction of cognitive effort during word recognition. For Trudeau (2003), advice given concerns the readability of the message, and not really its intelligibility. In other words, the guidelines give advice on surface structures, being influenced by classical readability studies that focus on lexical and syntactic levels (Nord, 2018), without taking into account the need to completely redesign the text. Moreover, guidelines sometimes tend to over-simplify (Labasse, 2006) and provide inconsistent advice with the literature. For example, asking for a "17-words meaning unit" or "a 17-words sentence" are two different things: the concepts of meaning unit and sentence do not entirely overlap. Finally, advice is hard to generalize, especially between the different areas of specialty; and even more so when adding new media and oral communications to which these guidelines are not suited (Nord, 2018).

Discourse and Writing/Rédactologie
Volume 31, 2021 http://journals.sfu.ca/dwr 55 However, these studies focus on guidelines from the early 2000s (or earlier), and do not necessarily take into account the latest advances in writing studies about communicational efficiency. They were also not designed for automation, and are mostly not based on the professional writers' characteristics or their actual practices. This is why we wanted to better map these practices via a survey.

Method
To investigate professional writer's practices, we conducted a cross-sectional survey, described in the first part of this section. We used a survey, as this tool is widely acknowledged to gather information about opinions or practices. We also chose to use an online survey to reach a wider audience that would not have been able or willing to travel for an interview, and to ensure the anonymity of the results. We collected the answers of 55 professional writers in Quebec. 1 As we said earlier, we chose to interview professional writers rather than functional writers given the importance of the writing and simplifying tasks in their work, and Quebec for its important place in the French-speaking area regarding the training of professional writers. The profiles of those 55 participants will be presented in the second point of this section. Lastly, we will discuss the analysis of the data collected, a combination of qualitative and quantitative approaches.

Research design
Our study is designed as a cross-sectional survey using a two-part questionnaire on LimeSurvey.
The first part focused on plain language practices of professional writers: their feelings about their work, their training, their views vis-a-vis criticisms often made against plain language, the limits of the practice, the assistance they used (especially plain language guidelines), but also what they would like to be provided with as additional help. All the questions were based on the research we presented in the "studies of writers' practices" section. We collected the opinions of the writers in the form of Likert scales, ranging from "strongly disagree" to "strongly agree", for 35 items, distributed according to aforementioned themes; nine multiple choice questions were also presented, when a scale was not appropriate. For example, the question "Have you ever used a plain language guide ?" and the section on plain language (only accessible for the writers who have already used guidelines) included closed-ended questions.
The second part of the questionnaire included a simplification task. The writers received excerpts from authentic texts (communications from a French-speaking administration to its Discourse and Writing/Rédactologie Volume 31, 2021 http://journals.sfu.ca/dwr 56 citizens) and had to choose whether or not they would adapt each of them. If they wanted to simplify an excerpt, they were asked to justify what they would change in the form of a list of the problems found in the extract or by rewriting the passage, or both (see Figure 1).
For this rewriting step, we chose to give writers only excerpts with difficulties. To ensure the proper representation of various types of difficulty, we selected them based on the literature on plain language, especially plain language guidelines.

Syntactic features
→ avoid long sentences (15 words per sentence) (1) → avoid prepositions which make the syntax of a sentence more complex and therefore harder to read (2) → make a separate sentence for each idea (2) Excerpt 4 (several sentences) (1) Links laws too frequently and without explanation (2) Use of unexplained acronym (its meaning must be inferred from another acronym)

Relational aspects
→ adopt reader's point of view and not yours (especially when you need to explain a law) (1) Lexical features → acronyms and abbreviations: be careful! Make sure everyone understands them (2) Excerpt 5 (vertical list) (1) Use of complex medical terminology, without explanation (2) Too many parentheses Lexical features → use common and concrete words (1) Relational aspects → observe the reading mechanism (don't overuse parentheses) (2) Excerpt 6 (one sentence) (1) Horizontal list of six elements, itself including a second list of six elements (2) Two ideas in one sentence Visual aspects → present enumerations vertically (1) Syntactic features → make a separate sentence for each idea (2) Excerpt 7 (start of a letter) (1) Letter looks like a circular (the tone is set from the start: "I apologize for the impersonal tone of this letter, which for practical reasons takes the form of a circular") The last one is related to the relational efficiency, even though it only focuses on specific low-level linguistic phenomena, often called the focus on the reader.

Relational aspects
We have therefore sampled excerpts from these five categories to be as representative as possible of different linguistic phenomena often found in plain language guidelines. We chose excerpts from authentic communications by using the platform AMesure (François et al., 2020) to highlight complex phenomena falling into lexical and syntactic categories, and by manually identifying the other categories. We also manually verified that excerpts did not include any other major difficulty that the one we focused on. Table 1 presents the difficulties that can be found in each extract. The difficulty categories are in bold, along with the corresponding advice for each difficulty. It should be pointed out that some categories are better represented than others, due to online format limitations.

Participants
The survey was administered through snowball sampling. It was disseminated from the Groupe Rédiger, who then sent the survey to other colleagues or to other mailing lists. The choice of this research team as the "root" is motivated by its eminent position relative to communication problems between Quebec administrations and citizens (Solar & Boucher, 2008). Moreover, it is established at the Université Laval, one of the first to have designed a professional writing program . However, there are still some differences between writers regarding their professional status, years of practice, and the importance of plain language in their everyday work. The distribution of those variables in our panel is shown in Table 2.
In this table, it is surprising to see that the students in the first task have on average about six years of experience. This fact can be explained by several writers interviewed having returned to formal education. They were already simplifying texts before resuming their studies.
The "experienced" and "confirmed" levels are quite similar in terms of experience, and could be considered equivalent in the analyses of their practices.

Data analysis
Once the submissions had been extracted from LimeSurvey, we anonymized the data before running the analyses. We used a mixed approach. A quantitative approach was used when we had values on a Likert scale (i.e. the first part of the survey). A qualitative one was necessary to standardize the simplifications of the second part, to make the answers easier to compare quantitatively.
In the first part of the survey, we had mostly Likert scales. We used percentages to better perceive the relative importance of the different answers. We also had closed-ended questions about the guidelines. To analyze them, we listed the different guides mentioned for each writer. We inquired about each guide to class them in two types: "old" guides (concerned by the shortcomings reported in our review of the literature) and the "recent" ones (giving advice on the whole text, with awareness of relational efficiency).
In the simplification task, we proceeded in several steps. First, a distinction was drawn between cases where writers made changes and the ones where they did not. Figure 2 presents, for each writer (one per column, anonymized), the excerpts they simplified. Black squares represent cases the writer considered as complex, and white squares correspond to the extracts where the writer considered nothing needed to be changed. On average, each excerpt was simplified 36 times, and each writer estimated that 8 extracts out of the 9 were complex. The "extreme" writers, bringing the average down by simplifying only 5 excerpts out of 9, are writers 15, 19 and 38. We tried to look for similarities between writers, which could explain this tendency to simplify less, and to see if it is focused on a specific category. To do so, our second step was to analyze the simplifications proposed in regard of the five categories of simplifications we mentioned: visual, lexical, syntactic, structural (textual) and Discourse and Writing/Rédactologie Volume 31, 2021 http://journals.sfu.ca/dwr 61 relational. For example, "this letter is rude, we need to change that" and "the tone isn't appropriate" are both normalized in "relational: change the tone of the text". Thirdly, we split the normalized simplifications in three categories in regard of Table 1: the simplifications we expected and that were performed, those expected and not performed (in less than 50% of the cases) and the additional ones (that we were not expecting).
Once our qualitative answers were transformed in normalized data, we were able to carry out a quantitative analysis. We used several characteristics of the writers: years of experience, importance of simplification in their work, training in clear writing, use of plain languages guidelines (and the type of guide used) and confidence in their practice. We considered their effect on the simplifications made, according to the five categories (calculated as the sum of simplifications made by a writer for each category over the total of modifications for that category), for each writer. Since we had predominantly ordinal variables, we opted for the Spearman's rank correlation coefficient to calculate this effect, using SPSS. Results of these analyses are also shown in the next section.

Results
In this section, we will present the results of our analysis, organized according to our three research questions. The first part describes writers' feelings about their own practices; the second presents the different aids used by the writers and what aids they would find useful. Then, the third part reports the analysis of the simplification task, completed by 40 writers. Simplifications performed by the writers will be described, before being compared to those proposed by plain language guidelines. This section ends up with a fourth research question, that emerged during the analysis.
We noted that some writers simplify less than others. The last point thus tries to identify such profiles and explain their habits on the basis of the writers' characteristics.

Feeling about clear writing practices
The writers were asked about their feelings regarding three main subjects: the criticisms often made of plain language, their own view of plain language, and the limits imposed by their environment. Table 3 shows the main results.

Plain language criticism
Plain language is ugly 0 % As we can see in Table 3, the writers' answers confirmed what has been found in the literature.
Thus, the -sometimes arbitrary -criticisms often made against plain language, synthesized mostly by Kimble (1994Kimble ( , 2016, are swept away by the writers. Plain language is not a leveling down, or a simple reduction in vocabulary or sentence length.
On the contrary, when asked what plain language is for them, the writers respond that it is above all a way of guiding the reader, of grouping the information and of prioritizing it differently. The changes made to the texts are both visual and linguistic. As we will see later, the vision of plain language has been confirmed in the simplification task. The modifications made to the lexical and Writers still remained well aware of the current limitations in the field. The job takes time to be done well and requires mutual help. Nevertheless, the writers interviewed seemed to have little direct concern with the limits presented in the literature. They feel supported by their colleagues, manage to target the recipient of messages, have received good training and do not feel limited in their practice. This may be due to the positive sphere of the Groupe Rédiger, and its desire to advance research in clear writing.
Finally, and before we focus further on the help given to writers, a quick word on the current simplifications in administrations. We asked whether administrative texts intended for a general audience were clear enough. The vast majority of writers (93%) say that they are not. For those who answered "yes", it was not a plain statement, but rather a positive note: simplification is improving, but there is still work to be done. Among those who answered "no", some writers explained why it was not yet clear enough. The 2 main reasons mentioned are: (1) that texts are not focused enough on the readers' needs (but on the needs of the administration), and (2) that legal content is not sufficiently popularized. Nevertheless, the writers agree that things are improving.

Help available to writers: assessments and prospects
To improve their clear writing practice, professional writers rely on two types of assistance: the knowledge and skills of their colleagues, and the various linguistic resources available. Table 4 shows the percentage of use of these various aids by our 55 respondents. said they are not under pressure from their superiors, and 15% of "neutral / no opinion"), they feel supported by their colleagues, and quite confident in their practice (87%, see Table 3).
As for external assistance, dictionaries are used by all writers. Searching for synonyms seems to be one of the most complex tasks to perform without the need for outside help.
Plain language guidelines are less used than any other type of external resource (excluding their own simplified texts). We investigated the reasons for this lack of interest and identified three main reasons. The first is ignorance of the existence of such a tool; one writer even argued that such aids could not exist. The second concerns the type of text they work on. The guidelines are not suitable for all types of texts (e.g. for social network posts or certain types of websites). The third reason is rather ideological, the writers privileging the knowledge acquired by practice to guidelines they consider "incomplete". This last category of writers could use a guide but prefer not to.
The writers who had already used plain language guidelines were asked to name the guides they used and provide their opinion on these. Although the literature often presents guidelines as incomplete or rather vague, and difficult to generalize (Nord, 2018), our writers contradict these comments. Indeed, apart from a lack of advice on the complete redesign of the text (accepted by almost 30% of the writers), more than 80% of the writers considered the advice as practicable. Two main hypotheses could explain this rather surprising phenomenon. On the one hand, the writers are professionals. They therefore have a good knowledge of the strategies to be used and are able to infer precise and useful information from the advice provided in these guidelines (which may not be the case for functional writers). On the other hand, most guides cited by the writers are from a "new generation", like De la lettre à la page Web (Clerc et al., 2006). These present advice on the whole text instead of a condensed list of rules like in older guides. However, the data collected in this survey do not allow us to choose between these two hypotheses.
When it comes to additional aid, 64% of writers favor a one-time help interface, via a website.
They would prefer this solution to an electronic guide (56%) or even to a program integrated into their word processor constantly evaluating the changes made (40%). This interface should highlight difficult passages -at lexical (72.5%), syntactic (75%) and semantic (80%) levels -and suggest alternatives -synonyms (80%), simplification of complex sentences (65%), reminder of plain language advice from a guide (55%). Even if this tool would be inspired by studies in automatic text simplification, it should not directly simplify the text (the only choice rejected by a majority of writers, 95%) but only propose solutions to the writer.

Types of simplifications performed
All excerpts proposed in the simplification task contained at least one linguistic difficulty recognized to alter comprehension. As explained in the Method section, we have chosen these extracts on the basis of a five-category typology to ensure the diversity of the problems and we have grouped the simplifications made by the writers into three categories: expected and carried out, expected and not always carried out (in less than 50%) and additional simplifications (see Table 5). 1) Explain undefined acronyms 2) Remove administrative and legal "jargon": references to laws, typical formulations of this kind of texts ("vous mettre en rapport", "connaitre à temps" -"get in touch", "know in time", etc.)

1) Simplify specialized terms from medical domain (excerpt 5)
2) Change of a generic term to avoid a reference problem (excerpt 8)

Syntactic features
1) Reduce sentence length: cut the sentence, rewrite it in several sentences if the sentence includes several ideas or is syntactically complex 2) Remove unnecessary clauses and passives 3) Switch negative, passive and impersonal turns to positive, active voice 1) Make one idea per sentence (lost in the vertical list) (excerpt 6) Structural features 1) Modify the example to add coherence to the story (excerpt 9) 1) Adapt the text to its channel: excerpts from websites don't necessarily respect the rules of this channel

Visual aspects 1) Switch from horizontal enumerations to vertical lists or tables for complex information
Relational aspects 1) Turn sentences in a personal way: as much as possible, speak directly to the reader, using "you" 1) Remove unnecessary parentheses (except 5) 1) Be considerate of readers: change the tone of the text, add politeness 2) Add additional information: in order to better understand the context, to facilitate contact between administration and citizens (adding an email address, a contact number)

Discourse and Writing/Rédactologie
Volume 31, 2021 http://journals.sfu.ca/dwr 66 All levels are represented in the "expected and carried out", except one (the structure of the text).
This is a surprising omission, but with a single excerpt it is difficult to highlight a trend. However, as we will see below, the structure is sometimes modified in other extracts.
Excerpt 5 is the only extract where the two tips from the guidelines are found in the "expected and not made" column. However, the writers always presented at least one of the two modifications, or a complete change of the text. This shows that, when there are many changes to be made, the attention of the professional writer is focused on the overall understanding, and not on more specific phenomena. The same tendency is found for excerpt 6, where the two ideas in a sentence are lost in the horizontal enumeration.
Additional simplifications have a common characteristic: they concern the whole excerpt. They do not focus on a particular linguistic phenomenon, but rather on the complete revision of the sentence or the extract. We also see the appearance of the notion of "considerate" writing, related to relational efficiency presented in the section "previous studies". In this last column, we can also add the text format changes (questions and answers, comic strip, etc.), proposed by one or two writers for some extracts.
We wondered if those differences (simplifications not made or additional changes) could be in connection with the plain language guidelines known by the writers. Some more recent guides advocate a complete revision of the text, rather than focusing solely on "classic" linguistic phenomena (as highlighted in Table 1). Unfortunately, the changes are mostly occasional and do not seem to follow the logic of a guide. They more probably reflect personal preferences of the writers, corresponding to their vision of plain language (as presented in the point "Feeling about clear writing practices"). Nevertheless simplifications presented here show the importance of the writer's know-how. They modify an important aspect of the texts that is missing in the guidelines: comprehensive relational efficiency (as shown in Method section). The simplifications tend to be more global, to relate to the general feeling given by the text, rather than to a specific linguistic phenomenon.

Writer profiles simplifying less
This section investigates the relationship between the writers profiles and the number of simplifications they made. Although all excerpts of the simplification task presented difficulties, only one passage was seen as difficult for all writers (see Figure 2). One example in particular Discourse and Writing/Rédactologie Volume 31, 2021 http://journals.sfu.ca/dwr 67 (excerpt 1) was even rated as "simple" by 20% of the writers. In this section, we try to identify the writers' characteristics that could explain why they did not simplify all extracts.
As we presented in the Method section, we performed a correlational analysis between six sociodemographic characteristics and the simplifications made by the writers. Results can be found in Table 6. As we can see in Table 6, three of the six characteristics have a significant effect on the simplification categories: importance of simplification in the work time, use of guidelines and type of guide used. The importance of simplification in the work time seems more important than the years of practice, the training or the confidence in the practice, showing that professionalism can be acquired in different ways (Gambier, 2016). It would be interesting to see the task made by functional writers who are also experienced. About the guidelines, as we saw in the previous point, some simplifications were not expected and are not included in classical guidelines.
So, we wondered if the correlations for these three significant variables would remain so if we distinguished between expected and additional simplifications. We believe that additional changes are more likely to be correlated to the writer profile. The results of the correlational analysis with the type of simplifications are shown in Table 7. As we can see in Table 7, the amount of simplification tasks in the work time has a significant effect on the presence of simplifications at the syntactic and relational levels. In both cases, the more experienced the writer is, the more they will make simplifications (syntactically, only for the expected items; relating to relational aspects, both expected and additional). The writers' habits can explain this: they can more easily put themselves in the readers' shoes and understand where the difficulties of the text lie.
The use of plain language guidelines also influences the relational aspects. Writers using guidelines are more likely to make simplifications relating to the reader's perspective. These writers also make more simplifications at the lexical level.
The type of plain language guide used has an influence on relational aspects, but also on the visual level. For relational aspects, only the additional changes are significant (see Figure 3 -upper line). Those using recent guides (such as De la lettre à la page Web) simplify more than those who use "older" guides; writers who do not use guidelines are the ones who simplified the least. For the visual aspects (as shown in Figure 3 -lower line), expected changes are more often made by the writers using older guides, while the additional changes are more likely to be made by writers using recent guides. This shows the evolution of the advice provided by these new guidelines, more focused on the needs of readers, and on other layout features (including visual codes related to the web area).

Discourse and Writing/Rédactologie
Volume 31, 2021 http://journals.sfu.ca/dwr 69 In conclusion, we can say that the variables significantly influencing the amount of simplifications at the lexical and visual levels are the use of guidelines and the type of guide used.
The most active writers are those using recent guides, and those who simplify the least are those who do not use any guide. The readers' consideration level is correlated with three variables: amount of work time spent on simplification tasks, the use of guidelines and the type of the guide.
This shows that a writer with simplification habits will focus more on this level. The use of guidelines (mostly the "new generation") can help the writer, in addition to the expertise gained with practice. In other words, practices seem to evolve not only with research on clear writing, but with the intrinsic questions of the writer (Clerc, 2008) as they gain expertise and good practices over the course of a career making simplifications. However, it should be noted that, in our panel of writers, some are more engaged in linguistic revision than in professional writing. This could explain why some respondents simplify less on the visual aspects, in particular.

Conclusion and perspectives
This study investigates the plain language practices of professional writers. We interviewed 55 professional writers in Quebec. Among these, 40 took part in a simplification task on excerpts from authentic texts produced by French-speaking administrations. We had three main research questions: professional writers' views about plain language (and limits of their current practices), what forms of assistance do they expect to have at their disposal and what are their actual clear writing practices.
We could see that a similar view of plain language is shared by all writers. Their main concern is to put the focus back on the reader and on its needs, by reorganizing the text and personalizing it more. This is the logic of recent writing studies, about communicational efficiency: "Write for your reader" (Beaudet et al., 2012), the relational efficiency (Romain et al, 2016), and so forth. "Classic" criticisms of plain language approaches have therefore been swept away; the clarity of a text is much more than a question of vocabulary or sentence length. Our panel of writers seemed aware, but not directly affected by the limitations of the environment. The main obstacles mentioned by the writers are pressure from clients and insufficient training in plain language.
There are two types of assistance to facilitate the work of writers: collaboration with other writers and use of linguistic resources. Our survey showed that dictionaries are used by everyone.
Plain language guidelines are the least used, despite being the main tool designed specifically for the task. Even though writers consider them to be easy to use (contrary to what the literature says), they see a problem in them: the lack of advice on a complete redesign of the text. The work of a professional writer aims not only to facilitate reading, but also to help the co-management of the text between client and reader.
In the simplification task, we observed once more this commitment not only to adjust the lexical and syntactic phenomena, but also to improve the relationship between the participants of the communication. This engagement becomes more important with the experience of the writer. By customizing the texts more, adding linguistic politeness, softening the tone, offering other formats, etc., the writers prove that the modifications made relate to many linguistic and communicational dimensions.
Let us note, however, some limitations of our study. By focusing on professional writers, we have ignored functional writers, who are arguably the most limited writers when it comes to simplification techniques. It would be good, in a next step, to carry out a similar survey with this population. Nor is the influence of the Groupe Rédiger within our sample trivial: all the writers Discourse and Writing/Rédactologie Volume 31, 2021 http://journals.sfu.ca/dwr 71 interviewed have a more or less indirect link with this team that is well-known for its cutting-edge research in plain writing. Indeed, we have seen that some writers who simplify less seem to be people who are more distant from this team and who have less knowledge of its research. This type of profile could be underrepresented in our study.
In order to better compare these first results and to analyze the presence of potential bias, we plan to carry out a similar survey with functional writers in Belgium. This country, which has hardly any academic training in professional writing, could offer us another vision of the application of plain language principles in the everyday work.
Another interesting path for research would be to look at the readers' use of those plain language texts, to see if some characteristics are more relevant to improve communicational efficiency than others. This would help the writers to know where to put their revision efforts in priority.
In conclusion, and in light of the writers' answers for both tasks, we can say that a plain language assistance tool would be welcome. It should focus on the lexical, syntactic and semantic levels, giving examples of simplification, but leaving the final decision to simplify or not to the writer. Such a tool would aim to draw the attention of writers to problems at more formal levels, thus allowing writers to focus on substantive and structural issues, thereby facilitating their daily work.