Introduction

1Although stylistics is known as the linguistic analysis of literary texts, it also includes analyzing non-literary texts. For instance, this very journal recently published works on journalism (Lacaze 2015; Mathurin 2017); communication at the 2009 climate summit (Percebois 2010); Lord Saville’s Bloody Sunday Inquiry (Barcat 2019); and various forms of American presidential discourse (Benoit à la Guillaume 2010, 2013, 2016, 2019; Bonnefille 2019; Boulin & Levy 2018). Elsewhere, the Journal of Language and Literature published by Sage for the Poetics and Linguistics Association (PALA) also includes some analyses of non-literary texts. As for recent books, Schubert and Werner’s edited volume on stylistics (2022) analyses pop culture, Rigg’s monograph (2022) analyses online news, and Ringrow and Pihlaja’s edited volume on stylistics (2020) analyses many forms of contemporary communication, including social media. All these developments justify defining stylistics more broadly than before.

2That said, student essays are one non-literary genre rarely studied in stylistics. In rhetoric and composition studies, scholars have analyzed student writing for years and have published their findings in journals such as Assessing Writing (an Elsevier journal), College Composition and Communication (a National Council of Teachers of English (NCTE) journal), and Journal of Writing Research (a University of Antwerp journal), to name just a few. They often focus on writing by students whose first language (L1) is English. Meanwhile, applied linguistic studies of writing by students whose second language (L2) is English appear regularly in Elsevier journals such as English for Specific Purposes, Journal of English for Academic Purposes, and the Journal of Second Language Writing, to name only a few. All this research reminds us that there is no shortage of student writing to study. As Aull (2020, 46) explains, starting in the 20^th century, rather than write merely personal essays, more and more university students had to write “research papers that relied on … finding external sources for support.” More specifically, as Aull (2020, 9) states, university students must often write two kinds of essay: “argumentative writing assignments, or thesis-driven tasks, in which showcasing persuasive reasoning is a primary rhetorical purpose, and explanatory writing, or analysis-driven tasks, in which showcasing interpretive reasoning is a primary rhetorical purpose.” While studies of student essays seem rare in stylistics, they clearly lend themselves to the kind of discourse analysis that is becoming more common in stylistics.

3In this study, I focus on a well-known hallmark of academic writing: citation integration, or using quotations from external sources. Mastering this skill is so hard that countless textbooks, such as Bailey’s (2018) or Cottrell’s (2019), devote many chapters to it. Acquiring the skill requires students to repeatedly take at least seven steps: find reliable sources to use; select appropriate passages to quote; present a debatable claim; introduce the quotation; present it to support the claim; reference it correctly; and end with a comment. This forms what Losh (2005, 86-89) called the “claim-evidence-warrant” framework. Losh (2005, 85-86) reduced the six parts of Toulmin’s (2003 [1957]) complex model of argumentation (with its notions of claim, qualifier, ground, warrant, backing, and rebuttal) to just three parts (claim, evidence, and warrant) so undergraduates could more easily understand them. A claim is a debatable statement, while academic evidence in the humanities is typically a direct quotation from another source, be it primary or secondary. After the quotation, a comment should follow which says a little more about the evidence.

4The “claim-evidence-warrant” pattern is less formally called a “quotation sandwich.” As Graff et al (2018, 47) state:

To adequately frame a quotation, you need to insert it into what we like to call a ‘quotation sandwich’, with the statement introducing it serving as the top slice of bread and the explanation following it serving as the bottom slice. The introductory or lead-in claims should explain who is speaking and set up what the quotation says; the follow-up statements should explain why you consider the quotation to be important and what you take it to say.

5While this analogy explains rather simply a complex skill in academic writing, quotation sandwiches really have four parts rather than three: a claim; a reporting clause and a quotation; and then a follow-up explanation. Graff et al mention “claims,” but not as the top slice of bread; and reporting clauses are the statements that “lead-in” to quotations. Yet because Graff and Birkenstein’s method offers students scaffolding (‘templates’) they can use, their undergraduate textbook remains highly popular in the US: They Say, I Say: The Moves that Matter in Academic Writing is now in its 5^th edition (Graff & Birkenstein 2021).

6American undergraduates with L1 English are the main target audience of They Say, I Say, but I recently started using it with my French undergraduates whose L2 is often (but not always) English. I wanted to find a solution to recurring problems of citation integration in their academic writing. An earlier study in our department catalogued grammatical errors in L2 English writing by 168 students in all three years of our BA program (Hamitouche 2020), whereas my study only focuses on quotation sandwiches in essays by 43 of our year 3 BA students. Results are mixed. At one end of the spectrum are a few students who master this skill; at the other end are a few more who show little mastery of it. In the middle are most students whose ability varies.

Background

7This is a study of essays by students who took my year 3 composition module, Anglais Ecrit, in both semesters in 2020-2021. Teaching academic writing is notoriously labor-intensive and time-consuming. Most writing teachers would agree with Thomas and Sassi (2011, 52) that “we have … an obligation to teach students the positive traits that help them avoid plagiarism, such as citing sources correctly.” Using templates may be a mechanistic way to do that, but it seems effective for many students. Therefore, in autumn my students completed exercises each week from They Say, I Say. Graff and Birkenstein (2018, 47) provide students with templates to use when introducing quotations:

1. X states, “Not all steroids should be banned from sports.”

2. As the prominent philosopher X puts it, “____.”

3. According to X, “________.”

4. X himself writes, “______.”

5. In her book, ____, X maintains that “____________.”

6. Writing in the journal Commentary, X complains that “____.”

7. In X’s view, “______.”

8. X agrees when she writes, “___.”

9. X disagrees when he writes, “______.”

10. X complicates matters further when she writes, “___.”

8They also provide templates for explaining quotations (Graff & Birkenstein 2018, 48):

1. Basically, X is warning us that the proposed solution will only make the problem worse.

2. In other words, X believes ___.

3. In making this comment, X urges us to ___.

4. X is corroborating the age-old adage that ___.

5. The essence of X’s argument is that ___.

9Students fill in the blanks on their own as the content is not the point (school uniforms, racism …). The point is giving students scaffolding so they quickly see how to make effective quotation sandwiches, and eventually, logical academic arguments. The “X” in each template encourages students to use integral citations. While the name of an author who is quoted appears in an “integral” citation, in a “non-integral” citation it does not (Swales 1990, 148; Hyland 1999, 344; Petrić 2007, 240). Using integral citations seems to be the norm for developing L2 writers (Lee et al 2019, 6). From the perspective of stylistics, integral citations represent direct speech (Simpson 2004, 31; Jeffries & McIntyre 2010, 88) for they tell us who said (or wrote) what.

10Using templates may not seem like an example of “liberation pedagogy” – Ann Berthoff’s (1990, 362) term for Paolo Freire’s work. In fact, Graff and Birkenstein continue a tradition from ancient rhetoric as well as more recent research, including Berthoff’s 1982 concept of “workhorse sentences” or Sharon Myers’ “moves” and “templates” from 2003 (Rhodes 2005, 135). Encouraging the correct use of direct, integral quotation might also limit patchwriting, which Click (2012, 47) calls:

a technique often used by students, NNES [non-native English speakers] and native English-speaking students alike, who have difficulty creating acceptable academic writing and/or comprehending the texts they are using as sources…. [It happens when] students select a relevant section of text from a source, and rearrange sentences, remove some phrases, choose replacement synonyms, etc.

11Students might patchwrite to try to avoid plagiarism, but it is a problem. For example, Neumann et al (2020) studied essays by 60 L2 English students at Concordia University in Montréal. Because they found that interventions to solve patchwriting were effective, they encourage EAP teachers who teach academic writing to focus on “paraphrasing and citation skills, solid understanding of source information, and work[ing] on source-specific language to discuss the ideas” (2020: 11). To help L2 English students improve their academic writing, one way to do this may be to use templates.

12While the previous paragraph shows a quotation sandwich in action, no matter how citation integration is taught, its importance cannot be ignored. In his study of 188 L2 English student essays at Masaryk University (Czechia), Docherty (2019, 38) argued that “both the form and function of direct quotation need to be taught ... for students to move more rapidly to a level of linguistic competency with regards to its use.” In other words, this skill matters. In their study of 100 ESL students at Ohio University, Lee et al (2019, 10) admit that while “source integration is but one dimension of successful academic writing, using citations effectively is crucially important in the dialogic construction of one's argumentation.” This is why they feel that “source and citation use … needs much more direct instructional attention … for L2 students than it may currently receive” (Lee et al 2019, 10). Petrić (2007) would agree. In her study of citations in 16 MA theses written in English, she found that using more citations and varying their formats affected evaluators positively. This is why Petrić (2007, 251) concluded that “source and citation use should receive more attention in EAP courses.” Plakans (2010, 188) reached the same conclusion in her study of “reading-to-write tasks” by 10 ESL undergraduates at the University of Iowa. She (Plakans 2010, 194) argued that “in academic English writing courses, students should be guided in developing a discourse synthesis writing process, including strategies to comprehend, select from, and integrate reading with writing. Providing exposure to writing that integrates other skills facilitates experience in understanding the demands of such writing for future academic coursework.” Textbooks like They Say, I Say contain such readings for this reason. And since L2 writers have trouble deciding how much material to quote from other sources (Lee at al 2019, 6), letting them study model texts also helps (Swales 1990; Dean 2010).

13After their exercises with sentences and paragraphs, my students adopted various They Say, I Say templates for their own research essays. The final essays for autumn and spring semesters were roughly similar. In the autumn they could pick their own topics (e.g., animal rights, inter-ethnic adoption, sport and politics, etc.), while in spring they had the whole semester to do their research on cyberbullying and social media addiction. They focused on those topics as that particular cohort was concerned about them. Their spring semester essays had titles such as the following: “Why are LGBT people bullied on social media?”; “The differences between bullying in school and cyberbullying”; “How does cyberbullying impact high school and university students?”; “Is social media making us anti-social?”; “How does social media affect children?”; “How does social media impact our everyday lives?”; “Why are some people racist on social media?”; and “The representation of masculinity in social media: the effects on society and individuals.”

14In this study, I analyzed the spring essays rather than the autumn ones because I assumed more skill would be visible in spring than in autumn. We must recall that this study took place in an English Department in France. The French Ministry for Higher Education and Research (MESR) recently established a list of skills (compétences) that BA students should master upon graduation. According to the MESR (2015, 37), upon graduation our students should officially be able to do many things, including “Identifier et sélectionner diverses ressources spécialisées pour documenter un sujet” (Identify and select various specialized resources to document a topic); “Analyser et synthétiser des données en vue de leur exploitation” (Analyze and synthesize data in order to use them); and “Développer une argumentation avec esprit critique” (Develop a critical argument). These skills are not discipline-specific, but I think that requiring undergraduates to do research and write essays may help students develop them. But why write argumentative essays? As Fang (2021, 146) writes:

An argumentative essay is a genre of writing where the author takes a position on an issue and provides reasoning and evidence to back it up. To make a credible or convincing argument, you need to collect and evaluate evidence, making sure that the pieces of evidence selected for inclusion – anecdotes, statistics, examples, quotes, testimonials, artifacts – are closely relevant and can bolster the argument you want to make. Claims or statements made without the support of evidence have little value, as they are simply opinions and not valid arguments from an academic perspective.

15The clear distinction Fang makes between “opinion” and “valid” academic argumentation is crucial for “quotes,” “support,” and “evidence” play big roles in these essays. Students who write research-based argumentative essays may come to see that, and thus move closer to the MESR goals.

16I must also add that formal L2 teaching in France involves the Common European Framework of Reference for language learning (CEFR in English; CECRL in French). The CEFR contains six skill levels (A1, A2, B1, B2, C1, C2), which range from A1 (the lowest) to C2 (the highest). When they graduate, most of our undergraduates are at or between B2 or C1 levels; few are beyond C1. Their receptive skills (listening and reading) tend to be better than their productive skills (speaking and writing). According to the Council of Europe (2020), B2 level writers can “synthesise information and arguments from a number of sources” (p.68); “develop a clear description or narrative, expanding and supporting their main points with relevant supporting detail and examples” (p.140); and “produce an essay or report, passing on information or giving reasons in support of or against a particular point of view” (p. 178). The Council of Europe (2020) also says that C1 level writers can “produce clear, well-structured texts of complex subjects, underlining the relevant salient issues, expanding and supporting points of view at some length …” (p.68); “expand and support the main points at some length with subsidiary points, reasons and relevant examples” (p.140); and “produce detailed expositions of complex subjects in an essay or a report, underlining … the salient issues” (p. 178). The MESR never says which CEFR level our majors must reach before graduating, yet the MESR targets (listed earlier) align with skills that the Council of Europe says B2 and C1 level writers should have. This is another reason why we ask our English majors to do research and write argumentative essays.

Methodology

17The study population had 45 undergraduates majoring in English in our English Department at the University of Upper Alsace (UHA) in France. None had English as their L1. They were in year 3 – the final year – of our BA program. As 2 students did not write a final essay, only 43 could be included in this study (35 women; 8 men). Of them, 28 had been in our university for no more than 3 years (65%), so they were on schedule by not having repeated a year; 11 had been there for 4 years (25%); and 5 had been there for 5 years or more (10%). In spring 2021, the students spent the entire semester doing research because “the more we know about a topic, the better we can write about it” (Dean 2010, 119). This is one reason why we require students to use references in their writing. For their final essays, students had to cite 5 to 7 good external sources, follow MLA style rules, write an essay containing at least 1,000 words, and have a classmate peer review it before they revised it and uploaded it to Moodle. Every week several students did short presentations on their bibliography in progress to show what sources they intended to cite in their final essays. At least 25 academic sources were also made available on Moodle so students could easily access current, high-quality, peer-reviewed research on gossip, social media, (cyber)bullying, and racism and discrimination. These spring semester essays are what Aull (2020, 9) calls “argumentative writing assignments, or thesis-driven tasks, in which showcasing persuasive reasoning is a primary rhetorical purpose.” Such argumentative essays based on research are common assignments in academic writing. As for using sources, Lee et al (2019: 5) found that the citation density of student essays tends to increase as they move up in the university system. They also note that direct quotations are commonly used by L2 writers of academic essays (Lee at al 2019, 6). Thus, having our year 3 students use templates and 5-7 sources in their essays seemed feasible.

18Table 1 gives an overview of the corpus. There were three steps to building the final corpus used in this study. First, students submitted their essays via Moodle in April 2021. Thus, they were not anonymous, but for this study they were analyzed one year after students had graduated.

Table 1. Study corpus

Items	Words
Corpus of 43 essays written in April 2021	51,667
Corpus of 39 essays included in the study	42,673
Corpus of 243 “quotation sandwiches”	19,805

19Second, 4 of the 43 essays were excluded from this study because they only contained paraphrases, no quotations. This left 39 essays for analysis. An example had to have quotation marks to be included; it did not matter if students used ‘British’ or “American” quotation marks. But without a quotation, a quotation sandwich is not a sandwich. Examples without any quotations marks at all were considered paraphrases, which I simply define here as citing external information without quotation marks. Such paraphrases might be plagiarism, which I did not focus on here, even if the Compilatio software program scanned all essays for plagiarism. In Click’s view (2012, 46), “unintentional plagiarism” is possible when students unknowingly attribute information from one source to another source due to sloppy research methods. This implies that intentional plagiarism occurs when students intentionally omit quotation marks around other people’s words. Without knowing what a student intended by apparently using paraphrases rather than direct quotations, I decided to exclude them from analysis. However, Petrić (2007, 243) included such items in her study of “citations” when they contained reporting clauses starting with “According to …” (which she called instances of “attribution”).

20Third, in the remaining 39 essays, the average essay was 1,094 words long and contained 6.23 quotations. In these 39 essays, there were 299 instances when students used outside evidence from external sources (243 quotations; 56 paraphrases). Just as 4 essays only containing paraphrases were excluded in step 2, the 56 paraphrases (4,568 words) were excluded in step 3. Ultimately, 243 quotation sandwiches totaling 19,805 words from 39 essays comprised the final corpus. Each example in context contained 81.5 words on average. By ‘context’ I mean a passage of 3 to 5 sentences containing a claim; a reporting clause and quotation; and then a comment/warrant.

Results

21While a quotation sandwich had to contain a quotation to be analyzed, Table 2 shows that 128 of 243 (53%) examples were correct.

Table 2. Generic results

Examples	Count
Correct	128 (53%)
Incorrect	115 (47%)
Total	243 (100%)

22A correct item had a claim; a reporting clause and quotation; and a follow-up comment which were all logically linked. Examples like (1) were considered correct in this study:

(1) This greed for the perfect body or the perfect behaviour is destructive for physical and mental health. In a study called “Psychology of men and masculinity,” the authors insist on the fact that, “The impact of ideal media images may also be seen in the increasing prevalence of eating disorder symptomatology, body dysmorphia, excessive exercise, and steroid use among men.” These images are not constructive, nor are they useful.

23In (1), the student first makes a claim (This greed for the perfect body or the perfect behaviour is destructive for physical and mental health). Second, the student correctly uses a reporting clause and an appropriate quotation (In a study called “Psychology of men and masculinity,” the authors insist on the fact that, “The impact of ideal media images may also be seen in the increasing prevalence of eating disorder symptomatology, body dysmorphia, excessive exercise, and steroid use among men.”). Third, the student ends with an appropriate comment on the evidence (These images are not constructive, nor are they useful). All three parts of the quotation sandwich are coherent, so (1) is deemed correct. Table 2 also shows that 115 of the 243 (47%) examples were incorrect. Because 3 of the 39 essays were found to have no errors at all, the 115 problems came from 36 essays in fact. As mentioned earlier, essays had 6.23 quotations on average; since 115 of them in 36 essays were incorrect, it means 3.19 problematic quotation sandwiches occurred per essay on average.

24As Table 3 shows, the total number of errors was 148 because 34 of the 115 (30%) problematic examples contained more than one error.

Table 3. Error total

Location of error	Number
In the claim	43 (29%)
In the reporting clause	45 (30%)
In the quotation	4 (3%)
In the follow-up comment	56 (38%)
Totals	148 (100%)

25As we see in Table 3, the fewest errors occurred in the claims (43), and most errors occurred in the warrants or follow-up comments (56). In the middle of those two pieces of bread in the quotation sandwiches are the reporting clauses and quotations (49). There we can see that students had far more problems in their reporting clauses (45) than in their quotations (4).

26Next, Table 4 shows more information about those 148 errors.

Table 4. Error type

Location of error	Number	Part missing	Incorrect part
In the claim	43	27 (63%)	16 (37%)
In the reporting clause	45	13 (29%)	32 (71%)
In the quotation	4	0 (0%)	4 (100%)
In the follow-up comment	56	34 (61%)	22 (39%)
Totals	148	74 (50%)	74 (50%)

27As Table 4 shows, half the errors (74) were due to the quotation sandwich missing a part, and half (74) the errors were due to them having an incorrect ingredient, as it were. The total error count is 148 rather than 115 because 34 of 115 quotation sandwiches contained more than one error. For example, in (2) at least two errors occur simultaneously:

(2) When thinking of Social Media, we should link the security of our data and how it can be misused or stored forever in a computer's memory. “If the social network account of one of the user's friends is hacked, the spammer or the hacker can misuse these details to blackmail the user” (qtd. in Aldhafferi 3). For instance, if someone manages to get our data by hacking the account, he could threaten the person's life by publishing photos or e-mails.

28In (2), on the face of it, there is a claim, a quotation, and a warrant, but no reporting clause. Instead, we see what Docherty (2019, 32) would call a “free-standing quotation” seemingly dropped into a passage without any reporting clause leading into the quote. Next, the quotation comes word for word from a journal article written in 2013 at New English University in Australia by Aldhafferi, Watson, and Sajeev. The attribution should thus be to “Aldhafferi et al 3” in line with MLA rules now about publications with three or more authors. But “qtd. in” means the words in the article by Aldhaferri et al. come from another source. What did Aldhafferi et al (3) actually write? They wrote: “If the social network account of one of the user’s friends is hacked, the spammer or the hacker can misuse these details to blackmail the user (Rosenblum 2007)”. They paraphrase rather than quote Rosenblum, but “qtd. in” makes it look like Aldhafferi et al are quoting another source. In this case, the reporting clause, quotation, and reference need to be fixed. Simply put, (2) shows how more than one error could occur in a single example.

Errors with claims

29As Table 4 shows, there were 43 errors with claims which were either missing (27) or irrelevant (16). In (3), rather than start the quotation sandwich with a claim, this student informally narrates:

(3) To give some numbers, I found on the website BLACKLINKO a long article, written by Brian Dean, giving a huge amount of statistics about social media. For example: “the average time a person spends on social media a day is 2 hours 24 minutes / 144 minutes.” (Dean) Moreover, a lot of people have accounts on these networks.

30In (3), Brian Dean’s words are presented in direct speech, yet the fact that the student found them online is not a debatable claim: it is a fact. Instead of starting with a claim, (3) starts with something like a “narrative report of action” (Leech & Short 1981, 324; Hoffmann 2017, 168) which is inappropriate for an academic essay. Meanwhile, in (4), the so-called claim is problematic:

(4) To explain where unrealistic expectations come from, a study about “Fitspiration and thinspiration” can help us out. In this study, the authors describe a concept: “Thinspiration, or inspirational messages promoting thinness, has received criticism for its detrimental effects on body image” (Alberga, A.S., Withnell, S.J ,M von Ranson, K). With this definition, we can therefore easily see where the problem lies.

31The student’s claim in (4) is that “unrealistic expectations” come from somewhere. The quote first defines “Thinspiration,” and then says it has “detrimental effects on body image.” The warrant then refers back to the “definition” while the reporting clause tells us “the authors describe a concept.” The student focuses more on this quote as defining a concept, and less on harmful messages, which is where the source says “unrealistic expectations come from.” A claim such as Harmful messages are the source of unrealistic body image expectations is one which the quotation in (4) would back.

Errors with reporting clauses

32As Table 4 shows, there were 45 errors in reporting clauses meant to introduce quotations. The errors usually fell into two categories. One type can be seen in (5):

(5) First of all, people share their information everywhere on networks like Facebook, Twitter, and where these networks make their money is through the advertising, they sell in the first place. “The primary way social media companies like Facebook and Twitter make money is through selling advertising.” (McFarlane). Consequently, people see advertisements in every Social Media they use, and that is how they make money, and sometimes their data is exploited too.

33Despite the reference at the end telling us that MacFarlane is the source of the quotation, the student simply drops the quotation into the passage without a reporting clause. Such “free-standing quotation[s]” (Docherty 2019, 32) occurred in 13 cases, but as 7 of them were from one single student, this problem was not so widespread. What I call dropped quotations or drive-by quotations are unsuccessful instances of what Hyland (1999, 344) called “integral” quotations in academic writing. From the perspective of stylistics, such errors may seem like examples of “free direct speech” (Simpson 2004, 32; Jeffries & McIntyre 2010, 88; Hoffmann 2017, 167). Thanks to the quotation marks, we know someone said something, but we do not know who.

34While omitting reporting clauses is one type of error, using an incorrect one is another, as (6) shows:

(6) In the same way, Martinescu highlights this role in her study published in the journal Elsevier, “gossip is a functional behaviour that enables individuals to exchange valued resources.” In other words, Martinescu supports Feinberg’s argument that gossip can improve social interactions and shape a unified society.

35The 2019 paper on gossip by Martinescu et al was published in the journal called Organizational Behavior and Human Decision Processes. The student ironically names it correctly in their works cited list, but in (6) confuses the journal (Organizational Behavior and Human Decision Processes) with its publisher (Elsevier). There were 32 errors like these in the corpus, so this problem was more widespread. It also included issues with embedded quotations (e.g. indicating correctly that a student is quoting someone a BBC reporter quoted first when the BBC is the student’s source).

Errors with quotations

36According to Table 4, just 4 pieces of evidence themselves seemed incorrect. After all, quotations had to be present to be included in the study corpus. Yet (7) shows a rare instance of an error in one:

(7) Cyberbullying is, nowadays, the main subject which is talked everyday, making some consequences on people's mind and body. Lawrence Robinson and Jeanne Segal write, “Being bullied can leave you feeling helpless, humiliated, depressed, or even suicidal. But there are ways to protect yourself or your child – at school and online – and deal with a bully” (HelpGuide). In other words, cyberbullying is aggressive behavior which is online.

37In (7), the claim and the warrant both emphasize cyberbullying, so they fit together. The claim mentions the “consequences” of cyberbullying, while the warrant clarifies what “cyberbullying” means. But the first half of the quotation is about the effects of all types of bullying, not just cyberbullying per se, while the second half seems like an introduction to solutions to solve the problem of bullying in all its forms, again not just cyberbullying per se. The claim also contains a grammatical error, but my study did not focus on such errors. Example (7) is what Docherty (2019, 32) would call a “shoehorned” quotation, which “placed into a sentence where it either does not fit, is not referred to directly or makes little to no sense at that position in the text.” Although there seemed to be few problems with quotations themselves in the corpus, in “shoehorned” cases the quotation itself appears to be problematic, rather than the bread around it in the quotation sandwich.

Errors with follow-up comments

38Finally, as Table 4 shows, the most frequent error occurred in follow-up comments (56 in total). Like some errors in claims and reporting clauses, a comment might simply be absent, as (8) shows:

(8) Secondly, some people are stepping in to fight racism on social networks. Indeed, fortunately, some people or institutions such as the government may be led to ask those responsible for applications to be more committed to the management of their applications, “in particular by systematically detecting and deleting racist messages”. (rtbf.be)

39Such omissions occurred 34 times (61% of 56). In other words, we see a quotation sandwich with just a single slice of bread and filling. In contrast, in 22 cases (39%) a follow-up comment was present, but it seemed incomplete or irrelevant, as (9) reveals:

(9) The article, “How Using Social Media Affects Teenagers,” published in child mind institute by Rachel Ehmke, shows that children exposed to social networks grow up with a lot of anxiety and less self-confidence. Based on a study conducted by the Royal Society for Public Health, the author said, “The survey results found that Snapchat, Facebook, Twitter, and Instagram all led to increased feelings of depression, anxiety, poor body image, and loneliness.” Not everything we see on social networks is accurate.

40In (9), the follow-up comment does not match the quotation. As the student says, “Not everything we see on social networks is accurate,” but the quotation does not discuss truth or accuracy. Rather, in Ehmke’s quote, she discussed bad effects from social media usage. Thus, the student has evidence that supports the claim, but an ineffective warrant at the end. This is better than having no warrant at all, and probably one of the easiest issues to resolve.

Discussion

41How effective are quotation sandwiches in our students’ essays? As a reminder, that was the question which drove this study. All things considered, the results tell a mixed story. On the one hand, 4 of 43 essays were excluded from the study because they contained no quotations sandwiches at all; and in another 5 essays, all the quotation sandwiches were incorrect. Thus, 9 of 43 students (21%) either could not or would not use quotation sandwiches, despite learning how to do so when writing similar essays in autumn 2020 with the They Say, I Say templates. On the other hand, 3 of the remaining 39 essays seemed error-free, and over half (53%) of the quotation sandwiches seemed to contain no errors. That is the good news. As for the 115 examples containing the 148 errors, the errors were equally divided between omissions (50%) and incorrect items (50%). As Table 3 reveals, 29% of errors occurred in claims (top slice of bread), 38% in the follow-up comments (bottom slice of bread), and 33% in either the reporting clause (30%) or quotation (3%) (i.e. the filling). Finally, as 34 of the 115 (30%) items had more than one error, so making one error sometimes led to another one being made, but not in the majority (70%) of cases. These are signs of hope.

42While 3 of 43 students wrote entirely correct quotation sandwiches, the remaining 40 students (93%) had varying degrees of ability with this skill. Counting errors in grammar or all errors in MLA formatting might have driven up the error rate detected here, but as Foster and Wiggleworth (2016, 98) remind us, the “point at which a learner can be said to have acquired a given feature of a second language (L2) is not something applied linguists agree upon.” In other words, success is hard to define. Is getting 75% of the quotation sandwiches correct in an essay good enough? Is getting most parts of each one in an essay good enough? As we begin to focus more on skill acquisition in France (l’approche par compétences), such questions may become more pressing.

43Academic writing takes time for students to learn because mastery does not occur overnight. As Artman et al (2010, 105) state, “Research skills cannot be taught in ‘one shot,’ just as writing cannot be taught in one term.” For this reason, our undergraduates take a composition module every single semester. “[T]he road to academic literacy,” according to Brent (2005, 258) “involves pedagogies of integration, extended process, and grounding in genuine inquiry.” That “road” takes years to travel. According to Ewert (2011, 14):

The development of reading and writing fluency in the first language is a long, incremental process that begins with oracy and continues over many years of schooling with countless implicit and explicit activities around texts. In the case of second language literacy development, the focus instead is often on explicit tasks that emphasize accuracy in comprehension and production of short decontextualized texts, which leads to disfluent reading and writing behaviors.

44Using templates and writing research-based argumentative essays can help L2 students avoid the problems with “disfluent” writing that Ewert mentions. Yet the skills our students need take time to learn in their L1, and even longer in their L2. This is another reason why my year 3 module focused on writing research-based argumentative essays. Students in year 3 who do not like them see that doing a research master degree may be the wrong choice for them.

45As for why errors occurred, I can only speculate here. I agree with Ellis and Barkhuizen (2006, 66) who state that “identifying the source of particular errors is not an easy task … [as] many errors are likely to be explicable in terms of multiple rather than single sources.” An error can occur for any reason, and some errors no doubt have many causes. Granted, I did no follow-up interviews, but even if I had, getting honest and reliable answers or explanations would be unlikely. “Why did you make a mistake?” is not a question any student can easily answer. It first requires knowing when a mistake has been made, but many students may not accurately estimate their ability (Dunning & Kruger 1999).

46At the very least, we know that many factors affect L2 learning (Saville-Troike 2012). First, by spring 2021, the Covid crisis had impacted half the study program of our year 3 BA students (the spring and autumn 2020 semesters, and the spring 2021 semester). It probably also affected their mental health, not only their L2 acquisition. Learning is hard, especially when everyone is suddenly online or in class wearing medical masks. Comparing essays from a pre-Covid semester to those from spring 2021 might therefore be interesting.

47Second, as an L1 can interfere with an L2, there could be a kind of ceiling effect (Saville-Troike 2012, 18). That is, if our students cannot already do something in French, doing it in English may be hard. Our students also take French composition classes each year, but they do not write the same kind of research essays there, nor is French always their L1. Persuasion in French differs from persuasion in English (Meyer 2015, 96), so the research-based argumentative essay may seem new to our English majors, even in year 3 in our department. Analyzing their French essays for citation integration in comparison might also be telling.

48Last but not least, thanks to the late Zoltán Dörnyei, we know that motivation plays a role in L2 learning. According to Dörnyei and Ushioda (2011, 9), “we humans are always multi-tasking in our day-to-day lives, and in the typical classroom students’ engagement in learning will interact with a complex variety of other competing attentional demands, activities, goals and pressures.” There is a war for attention in classrooms today: teachers want students to pay attention to one thing, but students often pay attention to something else (Smith 15). Our year 3 students in spring have 21 hours of classroom time each week; my module was but one of those hours. My module is in a language skills course worth 5 ECTS credit points and comprised of four parts: writing in English; speaking in English; translating texts from French to English (thème); and from English to French (version). So my mark was just 1 of 4 they got for the unit, and just 1 of 15 they got in spring 2021. Some students might have paid more attention to other modules, while others might have been bored or unmotivated by my task. The grading rubric I gave them in advance told them that they could get up to 5 points each for 4 different criteria: use of English; use of sources; organization; and content (including using the templates). Making effective quotation sandwiches may thus have seemed unimportant to some of them. As Brent (2005, 272) noted in his study at the University of Calgary, using citations may seem unimportant if it is too hard for students to grasp using them as a way of “leaving tracks” or “a trail of breadcrumbs” (Brent 2005, 273) for readers such that readers can easily follow a writer’s reasoning. In sum, the errors might be due to all these factors, either alone or in combination, and probably others, too.

Conclusion

49This small study is hardly pioneering as L2 error analyses are not new. There are also a few limitations that must be mentioned. First, my unit of analysis was the so-called quotation sandwich, typically comprised of 3 to 5 sentences in all. As Foster and Wigglesworth (2016, 102) note, delimiting “segmentation units” is challenging in any analysis. A researcher who used different units might obtain different results. Second, acceptability judgements are inherently subjective. As Ellis and Barkhuizen (2006, 56) state, “Acceptability is…dependent on the researcher…” Although they are referring to grammaticality judgements, my acceptability judgements are not entirely objective. Of course, having more than one rater would help, but unfortunately inter-rater reliability is never perfect either. Third, while this study contains some quantitative data, it is more qualitative than quantitative. Doing statistical analyses like those of Foster and Wiggleworth (2016) or Docherty (2019) might be interesting in the future. Such methods would make it possible to ask the corpus new questions and get new answers.

50The fact that 128 (53%) of the 243 quotation sandwiches were correct suggests the glass is half full. These are encouraging signs of learning in progress. Yet the fact that 20 (47%) of 43 students failed the assignment (with a grade of less than 10 out of 20) suggests the glass is also half empty. So what should a teacher do? Offering writing courses at university level which do not require students to cite sources correctly is not an option, nor will that help our students reach the official goals. When citation integration is a problem, ignoring it does students a disservice. Tellingly, in a study of 456 ESL students at Indiana University, most of them told Ewert (2011, 25) that learning how to write coherently, integrate quotations, and understand disciplinary conventions were all very important to them. Perhaps our students would say the same. And in her study of 75 ESL students at the University of British Columbia, Ling Shi (2011, 331) noted that it was misguided to imagine those ESL “students working their [own] way into a received view of appropriate source use.” Instead, this skill can be taught and learned. According to Foster and Wigglesworth (2016, 103), “as any teacher or researcher knows, there is more than one way to correct errors, and some involve fewer changes than others.” For L2 students, using templates can be a good place to start writing better, although other means are available.