The Naked Truth About the Naked This: Investigating Grammatical Prescriptivism in Technical Communication

Researcher’s note (June 2026). Stefanie Wulff and I wrote this in 2014 because the field kept treating “the naked this” as a settled rule, and we wanted to test that rule against what students actually do, using corpus data on attended versus unattended this. What we found was that the supposedly wrong choice is frequent, patterned, and often rhetorically smart, not arbitrary error. That same instinct—measuring real language behavior rather than enforcing a handbook rule—now drives my work on AI in writing instruction and assessment, where the question is again what the data show versus what we assume students should do.

Abstract

The decision to follow the demonstrative this with a noun phrase is important to students’ writing development. Previous research has emphasized when students should not attend this rather than studying why students make the choice. Using a corpus-linguistic approach, we investigated 1,999 instances of (un)attended this in student technical and academic writing. High shares of unattended this were found in both text types as well as in original and revised drafts.

Keywords: (un)attended this, binary logistic regression, corpus linguistics, distinctive collexeme analysis, student technical writing

INTRODUCTION

The (un)attended this functions as either a free-standing pronoun (1) or as a determiner that attends a noun phrase (2)¹.

1. This is part of a TCQ article.
2. This example is part of a TCQ article.

To many, (un)attended this does not merit investigation. Woolever (2005) labeled the variant “The Naked This”—her reminder to never “leave this naked without its noun” (p. 470). Pfeiffer and Adkins (2010) claimed that the sentence, “He talked constantly about the project to be completed at the Olympics. This made his office-mates irritable” only confused readers, forcing them to ask, “This what?” (pp. 646–647).

Although some may consider correct usage of (un)attended this a definitive and therefore unnecessary discussion, we argue that student writers’ decision whether to follow a demonstrative with a noun phrase is important to their writing development. Wulff, Römer, and Swales’s (2012) examination of this variant in A-graded student academic writing revealed that 43% of their samples’ instances were cases of unattended this, suggesting that the discouraged grammatical choice constituted a frequent choice in high-quality student writing. These researchers also identified verbs significantly associated with the unattended demonstrative (e.g., “This is/means/leads…”), indicating that students did not arbitrarily decide to leave this unattended.

A similar investigation, then, is merited in technical communication, particularly in light of recent critiques of the textbooks used in our classrooms. Wolfe (2009) reported how many generalist technical communication texts included discussions of stylistic and organizational principles that ran counter to the standards that engineers practiced on the job. More important to the present study, two technical communication textbooks have been identified as contributors of general advice regarding (un)attended this (Wulff et al., 2012).

In this article, we use a corpus-linguistic approach to investigate instances of (un)attended this in student technical and academic writing. We designed this study to advance beyond the mostly anecdotal and functionalist research on (un)attended this and to offer technical communicators insights into how their students think about writing and how grammar instruction could be modified. Our analysis is based on 1,999 hits obtained from a corpus of writing contributed by 18 students in a 16-week, sophomore-level technical writing course. The results of our study begin to suggest how students might use (un)attended this in technical contexts compared with academic ones. We also provide materials for use in technical communication classrooms and writing centers, including a list of verbs significantly associated with (un)attended this and a definition that encompasses the variant’s options.

REVIEW OF TEXTBOOK ADVICE ON (UN)ATTENDED THIS

We begin with an analysis of how popular technical communication textbooks discuss (un)attended this. Similar analyses have been conducted to examine textbooks’ effectiveness on a holistic level (McKenna & Thomas, 1997; Miller, 1992) and of specific subject areas (Barker & Matveeva, 2006; Wolfe, 2009). Because a study outside the field implicated two technical communication textbooks as contributors of general advice on (un)attended this, we wanted to assess if these discussions were reflective of the other texts used in our classrooms.

Fourteen technical communication textbooks published between 2005 and 2012 were initially selected for this analysis. All these textbooks had a generalist focus, and they were selected based on their longevity, popularity, or use in Wolfe’s (2009) recent analysis. Three of these textbooks contained no reference to (un)attended this and were removed from our analysis. The average textbook examined was in its seventh edition, a measure that Wolfe equated to a text’s continued popularity in the classroom (p. 354). Appendix A lists the textbooks initially selected for this study and a reference to each text’s discussion of (un)attended this.

We analyzed the 11 textbooks using Mackiewicz’s framework (1999) for identifying prescriptivism in writing handbooks. She found that prescriptive entries were generally presented in one of two ways: either (a) as a statement of the rule that governed the traditional usage or (b) as a pejorative that labeled a nontraditional usage. Semiprescriptive entries were also generally presented in one of two ways: either (a) as a discussion of what expert writers did compared with what readers preferred or (b) as a discussion that distinguished between formal and informal writing. Finally, prescription-breaking entries included those that eschewed the above-stated criteria and contradicted traditional usage rules. We found that the textbooks’ discussion of (un)attended this could not always be categorized unambiguously; however, Mackiewicz’s framework provided the best means for assessing and reporting our findings.

(Un)attended this was primarily described in each text’s grammar section; however, we also found discussions in sections that addressed how to help nonnative English speakers (Oliu, Brusaw, & Alred, 2010) as well as sections on paragraph development and sentence construction (Raman & Sharma, 2005).

Prescriptive Entries

Five of the textbooks we surveyed simply stated the traditional usage rule for (un)attended this, qualifying these entries as prescriptive. These statements were commonly brief in length and phrased with definitive, homogenous language: Pronouns “must clearly” (Lannon, 2009, p. 217; Reep, 2010, p. 497) or “should clearly” (Oliu et al., 2010, p. 629) refer to the nouns they replace. Two textbooks included extended definitions that cautioned writers that using a pronoun alone or to begin a sentence could confuse readers (Anderson, 2010, p. 226; Kolin, 2010, p. 704). The subheadings that organized these discussions were phrased neutrally, including

• “Pronouns,” (Anderson, 2010)
• “Articles” (Oliu et al., 2010)
• “Pronoun Reference” (Reep, 2010)
• “Correct Use of Pronoun References in Sentences” (Kolin, 2010).

However, Lannon’s “Avoid Ambiguous Pronoun References” subheading suggested the pejorative label that Mackiewicz (1999) identified as a second classifier of prescriptivism.

Some concession should be made to pejoratives used to label (un)attended this because a misuse will result in a vague pronoun reference. However, the subheadings used to introduce some discussions often went beyond the “Vague Pronoun” label (Johnson-Sheehan, 2012), including “Rule 3: Avoid Using This as the Subject Unless a Noun Follows It” (Pfeiffer & Adkins, 2010); “Problems with This” (Riordan, 2005); and the arguably pejorative “The Naked This” (Woolever, 2005). Further examination revealed that several textbooks also used pejoratives to label the writers who chose to leave this unattended. Riordan wrote that “many inexact writers” began sentences with this followed immediately by a verb (p. 559). Pfeiffer and Adkins claimed that the vague use of this was a “common stylistic error” that almost always “reflects poor mechanical style” (p. 646). Similarly, Raman and Sharma (2005) advised against using unclear pronoun references to avoid responsibility (p. 204); moreover, they attributed the use of unattended this to an overuse of short sentences that “sound choppy and disjoined,” often resulting in “jerky and irritating writing” that gave the “impression of elementary writing (such as writing for children)” (p. 190).

Semiprescriptive Entries

As noted earlier, we found that the textbooks’ discussions of (un)attended this did not always fit into the mutually exclusive categories that Mackiewicz (1999) outlined. In fact, four of the texts that labeled (un)attended this with a pejorative also offered suggestions for revision: “To guard against ambiguity,” Woolever (2005) wrote, “it’s good practice to turn the pronoun into a demonstrative adjective by pairing it with a noun” (p. 470). Two textbooks suggested multiple options for correcting an unattended this. Pfeiffer and Adkins (2010) instructed writers to make a sentence’s subject concrete “either by adding a noun after this or by recasting the sentence” (pp. 646–647). Riordan (2005) suggested that writers try “inserting a noun after this—and so turn it into an adjective—or by combining the two sentences into one” (p. 559). Johnson-Sheehan (2012) offered the most general advice on unattended this ambiguity—“train yourself to avoid using ‘It is…’ and ‘This is…’ sentences”—however, he acknowledged usage variation: “Occasionally, these sentences are fine, but some writers rely on them too much. You are better off minimizing their use in your writing” (pp. A-7). Markel (2012), too, presented a definitive rule for unattended this but conceded that “in almost all cases [emphasis added], demonstrative pronouns should be followed by nouns, rather than stand alone in the sentence” (p. 219).

Examples in Entries

More than half the textbooks included examples of the un(attended) this. We identified three example types, each with a distinct purpose: (a) to illustrate a revision by adding a noun to attend (and clarify) the demonstrative, (b) to illustrate a revision by recasting the sentence, or (c) to illustrate a violation of the rule and offer no revision.

1. Original: “He deals with personnel problems in his work. This helps him in his personal life. Revision: He deals with personnel problems in his work. This experience helps him in his personal life” (Oliu et al., 2010, p. 630).
2. Original: “The park ranger was pleased to see the workers planting new trees and installing new benches. This will attract more tourists. Revision: The park ranger was pleased to see the workers planting new trees and installing new benches, because the new trees and benches will attract more tourists” (Kolin, 2010, p. 705).
3. Example: “We really had a great week. Our program review went well, and we made huge strides toward finishing the project. This is why we are taking all of you to lunch” (Johnson-Sheehan, 2012, p. A-7).

The effectiveness of these example types, however, is subjective. It could be argued that the proposed revision for (a) and (b) lack a context with which student writers could identify, (b) is wordy and thus violates other conventions of technical writing, and (c) is just an obvious misuse of (un)attended this. Likewise, none of the examples we found included an illustration of when unattended this might be acceptable or even preferred. The assumption then, at least according to the textbooks we analyzed, is that these instances do not exist.

The absence of broader example types has been a noted limitation of grammar instruction in technical communication. In his discussion of unattached clauses, Jordan (1999) noted that some technical writing teachers “blindly follow” prescribed rules rather than use them as general guidelines. These teachers then use “obviously incorrect examples to ‘prove’ the point without also presenting examples that are debatable and those that are acceptable” (pp. 88–89). Jordan continued that these prescriptive rules can provide the initial foundation for effective instruction, but “[w]hen taught and enforced dogmatically and without real linguistic knowledge or sensitivity, however, they can do more harm than good” (pp. 89–90). Connatser (2004) also noted the field’s prescriptive approaches to grammar instruction and advocated for a stronger awareness of linguistic knowledge. He proposed that technical editors use more “organic grammar,” an approach that reconciles the grammar we are taught in school with the grammar we acquired naturally. Connatser reminded technical communicators that overriding natural language order with prescriptive rules often impairs the readability of documents and impedes how readers process information (p. 265).

PREVIOUS RESEARCH ON (UN)ATTENDED THIS

Writing research studies outside technical communication have specifically focused on uses of (un)attended this but mainly through anecdotal or functionalist perspectives that provide recommendations of when writers should or should not attend this rather than studying how writers actually consider these options (e.g., Geisler, Kaufer, & Steinberg, 1985; Moskovit, 1983). However, two recent studies have used authentic language data to advance understanding of when writers choose the unattended this. Swales (2005) examined the variant in published research articles from 10 disciplines and found a higher percentage of unattended this instances than anticipated, ranging from a low of 25% in dentistry to a high of 56% in philosophy. He hypothesized that many of these pronouns were unattended because of the “syntactically and semantically simple” verbs with which they were associated, for example, “This is/may be…” (p. 13).

Wulff et al. (2012) extended Swales’s (2005) study by analyzing instances of (un)attended this in a corpus of student academic writing. The researchers extracted more than 5,000 instances of sentence-initial this from the Michigan Corpus of Upper Level Student Papers, a corpus of 830 A-graded student papers across 16 disciplines. Forty-three percent of all instances in this sample were cases of unattended this, suggesting that the discouraged grammatical choice constituted a frequent choice in high-quality student writing. This finding sharply opposed the prescriptive recommendations referenced in writing handbooks and textbooks. The results from this study also suggested a strong influence of the verb accompanying (un)attended this: Verbs like “be,” “mean,” and “lead” correlated with unattended this, while verbs such as “use,” “examine,” and “focus” were accompanied by a specific noun phrase. When examined in context, different metadiscoursal functions of (un)attended this emerged that were bolstered by these verb preferences. Sequences like “This is/means/leads…” functioned as textual markers that signaled upcoming commentary and interpretation. In contrast, sequences like “This study uses/examines/focuses…” were predominantly associated with structural outlines, process descriptions, and other similar sections common to academic writing. In sum, Wulff et al. concluded that writers’ choice for unattended this was neither coincidental nor wrong. This study furthermore revealed that this main effect was moderated to some extent by author-related variables like academic discipline, academic classification, native speaker status, and gender.

The present study seeks to examine similar metadiscoursal functions related to the use of (un)attended this in texts produced in a technical writing class. Our analysis is based on 1,999 hits obtained from a corpus of student writing.

METHODS

Corpus linguistics is a modern applied linguistics approach that uses computer-assisted techniques to facilitate large-scale empirical analyses. A principled collection of texts—the corpus—is compiled and investigated with quantitative and qualitative analyses. The quantitative results reveal patterns in the language features that writers use, exemplifying the most typical language choice for certain functions in certain contexts. The qualitative analysis provides interpretation of reasons for typical and unusual choices (Conrad, 2002).

Corpus linguistics has been referenced as a useful (and complementary) tool for empirically oriented writing researchers (Orr, 2006; Römer & Wulff, 2010); however, only a handful of corpus-linguistic applications have appeared in our field’s journals. In a special issue of IEEE Transactions on Professional Communication, Orr noted numerous ways that linguistic approaches could be infused into technical communication scholarship. In particular, he cited corpus linguistics as a valuable means of exploring issues related to workplace communication. Corpus linguistics encompasses the study of human language through a representative sample of texts that allows researchers a depth of empirical knowledge that cannot be obtained through other field methods. In addition to responding to Orr’s call for more corpus-linguistic approaches to technical communication questions, we offer the first study that empirically and quantitatively explores (un)attended this in student technical writing. The design of our study provides a replicable model for investigating a variety of grammatical variants in our students’ writing, and the results offer unique insight into the decision-making process of our students and possible strategies for enhancing classroom learning.

Data and Variables

Eighteen students contributed to our study’s corpus. These students were all nonmajors and enrolled in an introductory sophomore-level course in technical writing. All data for this study were collected under exempt status (IRB #100288). The corpus included all the student writing produced over the 16-week semester, including the original and revised versions of each assignment. Students submitted the original texts throughout the semester and then submitted a portfolio of all their revised texts in Week 16. Each student contributed an average of 16 texts to the corpus (mean = 16, median = 18, SD = 3), accounting for 290 of 300 texts. The remaining 10 texts were written by small groups of the same students. Appendix B provides further demographic data on the students. The assignments consisted of technical texts—job materials, correspondence, procedures, and proposals—and academic texts—reflections on each instructional unit, a final reflection on the entire course, and reports.² Grade information was not collected from this sample; however, the inclusion of both original and revised texts arguably makes our sample a better representation of student writing than a more selective collection of texts that have passed a grade filter. Students were assigned readings from the Markel (2012) textbook, which included a semiprescriptive discussion of (un)attended this. Students received no formal classroom instruction on (un)attended this, which expectedly renders their use of the variant uncensored.

Data collection began with a case-insensitive search of all attestations (or hits) of this using AntConc, a freeware, multiplatform tool for conducting corpus linguistics research and teaching (Anthony, 2006, 2011). Each of the 1,999 attestations was then manually coded for nine variables: “attended,” “academic major,” “distinctiveness,” “draft,” “gender,” “language,” “position,” “text classification,” and “year.” Table 1 provides a description of each variable. The annotated data sample was subjected to two forms of analysis: a distinctive collexeme analysis (DCA) and a binary logistic regression. DCA identified specific verbs associated with (un)attended this. The binary logistic regression determined which, if any, of the variables listed above (or their interactions) discriminated significantly between instances of (un)attended this. We further describe both analyses in the following section.

TABLE 1 Variables and Variable Levels Considered in the Present Study

Variable	Description of Variable
Attended	Whether this was attended or not attended
Academic major	Student’s major in to two broad categories—STEM majors included students in applied economics, anthropology, chemistry, political science, psychology, and speech and hearing sciences. NonSTEM majors included students in human development and family studies, interior design, personal financial planning, retailing, and special education.
Distinctiveness	Value of the verb in the verb phrase most closely syntactically related to this in a given sentence
Draft	Each text as the student’s first submission or a revised submission
Gender	Student as female or male
Language	Language proficiency of the student as a native or nonnative English speaker
Position	Position of this within the sentence— • sentence-initial (i.e., when this was either the first word of the sentence, as in “This paper examines…,” or at least part of the first constituent of the sentence, as in “In this paper, we investigate.”) • sentence-medial (i.e., when this occurred somewhere in the middle of the sentence, and was neither part of the first nor last constituent, as in “I don’t know whether this is a good idea”) • sentence-final (i.e., when this was the last word of the sentence, as in “I like this,” or at least part of the sentence-final constituent, as in “I don’t like this ice cream”)
Text classification	Text type of the document as technical (i.e., job materials, correspondence, procedures, proposals) or academic (i.e., reflections, reports)
Year	Academic year of students (i.e., freshman, sophomore, junior, or senior)

Distinctive Collexeme Analysis

DCA is one variant of a set of corpus-linguistic methods called collostructional analysis developed by Stefanowitsch and Gries (2003) and Gries and Stefanowitsch (2004). All variants of collostructional analysis measure the association strength between words (typically verbs, as in our study) and the constructions in which they occur (e.g., [un]attended this in our study). Collostructional analysis is based on a constructionist perspective on language in which any pairing of form and meaning or function is defined as a construction (Goldberg, 1995, 2006). According to this definition, we can expect that highly conventionalized grammatical structures, such as (un)attended this, would have robust meanings or functions associated with them. The verbs statistically associated with a given construction are seen as a proxy to the meanings or functions a construction carries. In essence, this view reiterates the established idea in corpus linguistics that the meaning or function of a given word is captured in large parts in the words surrounding it, that is, its collocations—as Firth (1957) explained, “You shall know a word by the company it keeps” (p. 11). In accordance with a constructionist view of language, collostructional analysis expands this idea to pairings of words and the higher order grammatical constructions in which those words occur: The meaning or function of a grammatical construction should become transparent by the verbs that frequently occur in it (hence the term “collostructions,” which is a blend of “construction” and “collocation”).

DCA is a variant of collostructional analysis specifically tailored to identify verbs that distinguish best between two (or more) closely related or synonymous constructions. In the present study, we use DCA to address what verb lemmas, if any, are most distinctively associated with (un)attended this when we directly compare their frequency of occurrence in these two constructions. Technically speaking, to determine the distinctive association strength between a verb (X) and its two constructions (A and B), we need to obtain the following frequency information:

• the observed and expected frequencies with which the verb X occurs in construction A (n_{obs X in A} = n_{exp X in A})
• the observed and expected frequencies with which the verb X occurs in construction B (n_{obs X in B} = n_{exp X in B})
• the observed and expected frequencies with which other verbs occur in construction A (n_{obs ¬X in A} = n_{exp ¬X in A})
• the observed and expected frequencies with which other verbs occur in construction B (n_{obs ¬X in B} = n_{exp ¬X in B})

These frequency values are entered into a 2-by-2 table (Table 2).

TABLE 2 Frequency Values Entering a DCA of a Verb X

	Verb X	Other Verbs	Row Totals
Construction A	nobs X in A (nexp X in A)	nobs ¬X in A (nexp ¬X in A)	nA
Construction B	nobs X in B (nexp X in B)	nobs ¬X in B (nexp ¬X in B)	nB
Column totals	nX	n¬X	nsample

Table 3 is an example of what Table 2 looks like for the verb “be” in our data sample. In our data sample, “be” occurs 578 times overall: 317 times with attended this and 261 times with unattended this. These numbers alone suggest that “be” is somewhat biased toward attended this. The DCA, however, qualifies this observation in two ways: (a) it compares these observed frequencies with the frequencies we would expect if “be” were entirely randomly distributed across (un)attended this; and (b) it accounts for how many other verbs occur with (un)attended this. As to the latter, because we know the overall frequencies of attended and unattended this in our sample are 1,331 and 668, respectively, we can deduce that 1,014 instances of attended this and 407 instances of unattended this are accompanied by verbs other than “be.” As to the former, once we have determined all observed frequencies, we can calculate the expected frequencies of “be” (as well as other verbs in the sample) by multiplying the row and column total of each cell and dividing it by the entire total of 1,999. As Table 3 shows, if the distribution of the 578 instantiations of “be” were indeed to reflect the general frequencies of attended and unattended this, we would expect “be” to occur 385 times (1,331 × 578 = 1,999) with attended this and 193 times (668 × 578 = 1,999) with unattended this. In other words, we would expect “be” to occur with attended this more often than it actually does; conversely, we would expect “be” to occur less frequently with unattended this than we observe. A DCA captures the degree of deviation of observed and expected frequencies in the form of an association measure. The association measure we chose for the present study is a Fisher Yates exact p-value,³ which we report in its logged form for ease of interpretation. Absolute p_log values equal to or higher than 1.3 are significant at the 5% level of significance; positive p_log values indicate a distinctive association of a given verb with attended this, and negative p_log values indicate that a given verb is distinctively associated with unattended this. Correspondingly, “be” yields a p_log value of −11.769, which means that it is highly distinctively associated with unattended this. In collostructional parlance, that renders “be” a “collexeme” (a blend of “collocate” and “lexeme”) of unattended this when considered in direct contrast to attended this.

TABLE 3 Observed (and Expected) Frequencies of Be With (Un)Attended This

	Be	Other Verbs	Row Total
Attended this	317 (385)	1,014 (946)	1,331
Unattended this	261 (193)	407 (475)	668
Column total	578	1,421	1,999

For the present study, we used Gries’s (2007) coll.analysis 3.2, an R-script that allows the user to load all relevant data (in our case, the frequencies that all attested verbs occurred with either attended or unattended this), and the script returns p_log values for all verbs and information about which of the constructions they are distinctively associated with.

Binary Logistic Regression

We also computed a logistic regression analysis (a) to identify the most influential determinants of the distribution of (un)attended this, (b) to see whether, and to what extent, the choice of attended or unattended this was indeed associated with the linkage between either construction and the verb it occurs with as suggested by the DCA, and (c) to identify possible interactions of determinants that would remain undiscovered by a monofactorial approach.

Technically speaking, a binary logistic regression determines the probability of an event that can surface in two distinct ways. In our case, we want to determine the probability of the predicted level of the dependent variable “attended” on the basis of the following independent variables: “academic major,” “draft,” “gender,” “language,” “position,” “text classification,” and “year.”

A logistic regression works as follows: In a first step, all predictors (i.e., variables and their potential two-way interactions) are entered into a logistic model. On the basis of model comparisons, the predictor (starting from the highest level of interactions) that makes the least significant contribution to the model is identified and discarded, and another logistic model is computed without this predictor. This model-fitting process is performed iteratively until only significant predictors remain in a final model, which is also referred to as a minimal adequate model. More precisely, the minimal adequate model will contain significant predictors and nonsignificant predictors as long as the latter are part of a significant interaction.⁴ We present a minimal adequate model based on our data sample in the following section.

RESULTS AND DISCUSSION

Frequencies of (Un)attended This across Independent Variables

Out of the total data sample of 1,999 hits, 1,331 (67%) were instances of attended this, and 668 (33%) were instances of unattended this. Although shares of attended this were slightly higher than the ones reported by Wulff et al. (2012) for academic writing (57% versus 43%), we also found that shares of unattended this were far from marginal in the present data sample. The results of the logistic regression to be presented below will clarify where exactly (i.e., text classification, position) these instances of unattended this occurred. Table 4 provides a descriptive breakdown of the distribution of all 1,999 hits across the independent variables considered in this study.

We summarize Table 4 as follows: Overall, the data sample comprised about equal shares of instances authored by students majoring in science, technology, engineering, or mathematics (STEM; 51%) and nonSTEM (49%) disciplines. The sample contained twice as many original drafts (67%) than revisions (33%). Female writers outnumbered male writers by a ratio of 3:1 (73% to 27%). Only 5% of the instances in this data sample were contributed by a nonnative speaker. As far as the position of (un)attended this is concerned, 52% of instances occurred sentence initially, 24% occurred in a medial position, and 14% occurred in the final position. Sixty-three percent of the instances were classified as academic; 37% were technical. Lastly, 60% of the student writers contributing to the data sample were seniors; juniors and freshmen/sophomores contributed 18% and 22% of the hits, respectively.

TABLE 4 Distribution of Instances of (Un)Attended This by Independent Variables (n_{data sample} = 1,999)

Independent Variable	Variable Level	Instance	Percentage
Academic major	Nonstem	1,017	51
	Stem	982	49
Draft	Original	1,337	67
	Revision	662	33
Gender	Female	1,468	73
	Male	531	27
Language	Native	1,906	95
	Nonnative	93	5
Position	Initial	1,044	52
	Medial	676	24
	Final	279	14
Text classification	Academic	1,264	63
	Technical	735	37
Year	Freshman/Sophomore	361	18
	Junior	448	22
	Senior	1,190	60

Distinctive Collexeme Analysis

The DCA revealed a variety of verbs significantly distinctive for (un)attended this. As shown in Figure 1, verbs were sorted according to their distinctive association with attended this; correspondingly, verbs with positive p_log values significantly prefer attended this, whereas verbs with negative p_log values significantly prefer unattended this. (Appendix C provides the exact p_log values for each verb and its observed frequencies with [un]attended this.) Results of the binary logistic regression suggest that student writers are sensitive to these verb-specific preferences.

[Figure 1: Verbs distinctively associated with attended this (positive p_log values) and unattended this (negative p_log values). — image to be added.]

Binary Logistic Regression

The binary logistic regression determined which, if any, of the variables listed above (or their interactions) discriminated significantly between instances of (un)attended this. The minimal adequate model shows a highly significant correlation between the predictors listed in Table 5 and the choice of (un)attended this (log-likelihood ratio χ² = 418.2; df = 18; p = 0). Nagelkerke’s R², an indicator of general correlational strength between the predictors and the dependent variable, amounts to .262, and the model has near-good classificatory power: C = .77. (Usually, a C value of .8 or higher is considered “good” [compare with Harrell, 2011].) On the basis of the minimal adequate model, 72% of all instances can be predicted correctly as either attended or unattended this (the random classification accuracy amounts to 67%). Table 5 reports the significant predictors of the minimal adequate model in descending order of their absolute coefficient values as obtained from the lrm output.⁵ Positive coefficient values indicate a positive correlation with unattended this; negative coefficient values indicate a positive correlation with attended this. The dotted line in Table 5 separates predictors associated with attended this (top part) and unattended this (bottom part).

TABLE 5 Significant predictors of the minimal adequate logistic regression model

Predictor	Coefficient	p
POSITION = final	−2.097	0.000
POSITION = medial	−1.313	0.000
TEXT CLASSIFICATION = technical × DRAFT = revision	−1.111	0.000
GENDER = male × YEAR = senior	−0.898	0.005
YEAR = junior	−0.867	0.002
GENDER = male × YEAR = junior	−0.797	0.053
LANGUAGE = nonnative	−0.767	0.018
DRAFT = revision × YEAR = senior	−0.628	0.033
DISTINCTIVENESS	−0.088	0.000
…
POSITION = medial × YEAR = junior	1.461	0.001
DRAFT = revision	1.168	0.000
GENDER = male	0.678	0.013
YEAR = senior	0.432	0.045

“Draft,” “gender,” “position,” “text classification,” and “year” were identified as significant predictors that were also engaged in significant interactions. “Distinctiveness” and “language” were identified as main effects.

Interaction Between “Text Classification” and “Draft”

The interaction yielding the highest coefficient in the minimal adequate model was between “text classification” and “draft.” Both the technical texts and the revised drafts were generally associated with high shares of unattended this (see Figure 2, panel 1). However, we found significantly higher shares of attended this in revised technical texts; conversely, revised academic texts contained higher shares of unattended this than original drafts (see Figure 2, panel 2).

This finding might initially suggest that our student writers replaced instances of unattended with attended this in their revised texts. However, a qualitative analysis of the data reveals that these revisions were not commonly focused on correcting attended this; rather, students comprehensively rewrote the relevant passages for various reasons, with neither attended nor unattended this being selected. As examples, the revision in (f) appeared to be the result of additional research. The revision in (g) added the imperative mood, a common stylistic convention in instructions. Both examples seem to demonstrate students’ evolving understanding of technical writing fundamentals, rendering the choice between attended and unattended this obsolete.

[Figure 2: Interaction between text classification and draft. — image to be added.]

(f) Original: Teachers in Spain do not make as much money as U.S. teachers, but this is because of Spain’s higher level of unemployment.

Revision: Teachers in Spain make 22,964 a year compared to teachers in the U.S. who make only 21,787 per year (Education Indicators, p. 1). […] Also, Spain has higher levels of unemployment compared to the U.S. [multicultural redesign report, student H]

(g) Original: To start, you should prepare all of the ingredients necessary to present this to the customers and place it on a tray.

Revision: Place all necessary ingredients on a tray to present to the customers. [instructions, student C]

Interaction Between “Year” and “Gender”

The second significant interaction was between “year” and “gender.”

As illustrated in Figure 3, female students preferred to attend this rather than leave this unattended (numbers are particularly high for females in their junior year). Wulff et al. (2012) found a similar interaction between discipline and gender in their study of (un)attended this in academic writing, suggesting that our result from a smaller data sample is not an anomaly. Our gender effect aligns with the idea that females construct more formalized texts than males. In his study of the complementizer that, Jaeger (2010) noted that past research indicated that social variables correlated with syntactic choices (Adamson, 1992; Fries, 1940). Specifically, researchers have observed that female speakers are more likely to choose more formal registers than male speakers (Finegan & Biber, 2001). Attending the demonstrative, as our study’s female sample preferred, signifies the formal application of this variant and bolsters the previous findings on social variables and language.

[Figure 3: Interaction between year and gender. — image to be added.]

The male writers in our sample suggest a more complex picture. Freshmen and sophomores used unattended this twice as often as attended this; seniors used unattended this but less frequently than their male freshman/sophomore peers and the female students in their cohort. Junior-level males rarely left this unattended. These results suggest that, in particular, the male students’ use of unattended this fluctuates dramatically as a function of academic advancement (and arguably, writing proficiency). These particular results do not provide answers to several questions that emerge: Are junior students reacting to exposure to prescriptive usage rules (that may be less fresh in senior students’ minds or outbalanced by students’ increased confidence in their writing abilities)? Do junior students in our data sample use unattended this less often by sheer coincidence? Would we have found that these same juniors used unattended this more often in later stages of their college writing career?

”Language”

Although this study’s sample included only one nonnative speaker, “language” was one of the two significant main effects found. This result, although statistically significant, is not generalizable; nonetheless, it may encourage further research into native–nonnative writer differences in the use of (un)attended this and related phenomena. The English-native writers employed attended this 66% of the time; the nonnative writer preferred attended this significantly more, at 81% of the time. One possible explanation for this significant effect could relate to the prior English instruction that the student received. As language learning is more explicit and conscious than language acquisition, our L2 student may have received prescriptive instruction on demonstratives, which our earlier review of literature suggests is highly probable. The significant use of attended this, therefore, could reflect the difficulties nonnative speakers encounter when consciously adapting these prescriptive writing rules in ways their native speaker counterparts do implicitly. However, this explanation must be heavily hedged because we did not collect information on previous language instruction and because the findings from recent corpus linguistic research suggest that, when it comes to academic writing expertise, the L1-L2 differences are not always as pronounced as assumed (Römer, 2009a; Wulff & Gries, 2010; Wulff & Römer, 2009).

Interaction of “Draft” and “Year”

As we already mentioned, shares of unattended this were relatively high even in revised drafts (especially in academic texts), and shares of unattended this fluctuated based on a student’s academic classification. The observed interaction between “draft” and “year” begins to suggest how writing proficiency affects revisions. Figure 4 summarizes this interaction.

[Figure 4: Interaction between draft and year. — image to be added.]

The picture that emerges from Figure 4 is strikingly similar to the one observed earlier between “year” and “gender”: Freshmen and sophomores used unattended this frequently—28% of the time in original drafts and 57% of the time in revised drafts. In contrast, juniors generally avoided unattended this (although even in this group, we find that unattended this is more frequent in revised drafts than in originals). At the senior level, shares of unattended this are again higher (if only lower than freshman and sophomore students’ texts). The tendency to use unattended this more often in revised drafts than in original versions persists across all three student levels. This interaction seems to support our hypothesis that high-quality student writing does not equate to an absence of (un)attended this: Considerable shares of unattended this (a) stabilize in senior student writing and (b) are even more prominent in revised than in original drafts.

”Distinctiveness”

The second significant main effect observed was “distinctiveness,” which were the p_log values for each verb lemma obtained from the DCA. The directionality of the effect was as expected: The higher the p_log values (i.e., the more distinctive a verb lemma was for attended this), the less likely unattended this became (hence the negative coefficient in the logistic regression). Conversely, the student writers used verbs with highly negative p_log values (indicating a high association strength with unattended this) significantly more often with unattended this than with attended this. The box plot in Figure 5 summarizes this effect: The median p_log values of a verb’s occurring with attended this are significantly higher than that of verbs occurring with unattended this.

The right-hand box in Figure 5 is bigger than the left-hand box—this reproduces the fact we observed earlier that the amplitude of the p_log values of verbs distinctive for unattended this is considerably larger than that for verbs distinctive for attended this.

[Figure 5: Main effect of distinctiveness. — image to be added.]

The verbs “be” and “mean” appear to carry most of the weight here; both yielded p_log values lower than −10 (−13.008 and −11.769, respectively). This result accords with the Wulff et al.’s (2012) finding that “be” and “mean” (among others) strongly correlated with unattended this and in sentence-initial positions. These combinations of “be/mean” and unattended this perform a meta-discoursal function of summarizing, or commenting on, what has been said before. This finding resonates well with the Geisler et al. (1985) recommendation that instructors should allow writers to use unattended this when referring to the “central predication,” a term used to classify the most important action or event in the preceding sentence (pp. 146, 151–152). The following examples from our data illustrate this meta-discoursal function of initiating summary and commentary:

(h) Also women are not likely to work in the Kingdom of Bahrain. This was important in the redesign of my instruction set because it allowed me to narrow my targeted audience to younger males. [original draft of reflection, student J]

(i) The last of the cultural dimensions identified by Hofstede, long term orientation, does not contain a score for Spain. This is because originally, data was compiled for 70 countries. [revised draft of multicultural design report, student D]

(j) The assistant managers for each department will be put in charge of making all copies. This means that we need to turn in all the needed information to them. [revised draft of bad news memo, student M]

(k) Sorry for any inconvenience this might have caused… [revised draft of adjustment e-mail, student A]

Examples (h) and (i) exemplify the account of Geisler et al. (1985): The this in these sentences does not refer back to one specific referent but rather to the entire preceding predications (“Women are not likely to work in the Kingdom of Bahrain”/“The last of the cultural dimensions identified by Hofstede, long term orientation, does not contain a score for Spain”). The unattended this in (j) performs a slightly different rhetorical function. The scenario for this assignment featured the student as an assistant personnel manager who was faced with a photocopying budget deficit of several hundred dollars due to employees’ excessive use for personal reasons. The assignment required students to convey this budget news via a memo and then detail a plan to curb personal copy making. During this instructional unit, students studied guidelines for conveying bad news, including using the passive to avoid accusations and avoiding you-inclusive language to blame the reader. The student’s use of this in (j) then suggests that an unattended demonstrative could be an effective element to indirect writing style and an important marker for writers to learn when faced with composing a message their readers might react negatively to.

Though “be” and “mean” were the only verbs associated with the unattended this in academic writing, students used different, and arguably stronger, verbs (e.g., “allow,” “help,” and “cause”) to serve the same function in their technical writing. The use of “cause” in (k) is a particularly interesting example. One may wonder if the student is being deliberately vague in an attempt to veil the referent of this (i.e., a failure on the part of the company that spurred the customer complaint). Similar to example (j), a description of the adjustment e-mail assignment might identify an important rhetorical function of unattended this. The scenario for the assignment featured the student as a customer service representative for a software company. Students were provided with a legitimate complaint from a customer who reported a discrepancy with the promotional materials that advertised his new computer and the features that were actually installed on this computer. The assignment asked students to respond to the customer’s complaint via e-mail and either accept or reject his request. The text in (k) seems to illustrate the writer’s attempt to buffer the customer’s displeasure by using the unattended this to reference the entire complaint. The unattended demonstrative combined with the sentence’s sentiment (an apology) also demonstrates a sophisticated means of attempting goodwill with the customer despite an unpleasant shopping experience. Although the rhetorical situations surrounding (j) and (k) are different, the use of unattended this in both further suggests how unattended demonstrative pronouns might play a role in effectively delivering or responding to negative messages.

In addition to the summarizing function, we find that in student technical correspondence, sentence-initial “be” was used to stand in for the given letter/memo/e-mail and its purpose:

(l) This is to inform you of a new policy pertaining to the use of the photocopy machine. [original draft of bad news memo, student I]

(m) This is in response to your email regarding the upgrade of your computer software with recent graphics package, grammar check and multiple fonts. [revised draft of adjustment email, student I]

Interaction Between “Position” and “Year”

The final significant interaction was found between “position” and “year.” As Figure 6 shows, shares of unattended this were highest in sentence-initial position (especially in freshman/sophomore writing), considerably less frequent in sentence-medial position, and nonexistent in sentence-final position. However, the junior writers never used unattended this in sentence-initial position. The senior writers, in contrast, employed unattended this at the beginning of sentences nearly as often as they used attended this. Once again, we observe a sharp contrast in our sample’s junior students compared with their peers. We can only speculate about the reason for this effect; one possibility is that what we observe is a result of technical writing instruction cautioning against the use of unattended this.

[Figure 6: Interaction between position and year. — image to be added.]

PEDAGOGICAL IMPLICATIONS AND CONCLUSIONS

The results of our study can be summarized as follows: A total of 1,999 hits were identified in our data sample of student technical and academic writing: 1,331 (67%) were instances of attended this, and 668 (33%) were instances of unattended this. We used DCA to identify the specific verbs associated with (un)attended this (see Figure 1). The DCA identified several verbs as distinctively associated with (un)attended this; “be” and “mean” were the top two verbs distinctive for unattended this. Student writers used this + “be”/“mean” clusters to perform metadiscoursal functions of summarizing or commenting on previous statements (h–k). Our inclusion of student technical writing furthered the results of a previous study on academic writing (Wulff et al., 2012) by revealing that sentence-initial “be” also stands in for the purpose statement in technical correspondence (l–m).

Further, a binary logistic regression identified five significant predictors: “draft,” “gender,” “position,” “text classification,” and “year.” Perhaps the most surprising result of the study was the high shares of unattended this in student technical writing and in revised drafts. These results problematize the prescriptive discussions of (un)attended this in technical communication textbooks, specifically those texts that use pejoratives to label writers who choose to leave demonstratives unattended (Pfeiffer & Adkins, 2010; Raman & Sharma, 2005; Riordan, 2005). Geisler et al. (1985) made a similar observation concerning student writers’ use of the (un)attended this:

Good writers understand that readers rely on cues of focus to predict what’s coming next, and they also understand the importance of giving a reader good odds on his or her predictions. Student writers, on the other hand, typically force the reader to play with lower odds. They often do not control their use of the demonstrative, nor do they understand, even implicitly, the considerations of focus on which its use can depend. (p. 142)

The results of the present study contradict these assumptions and suggest that developing student writers’ use of (un)attended this can be rhetorically effective. Raman and Sharma (2005) advised against using unclear pronoun references to avoid responsibility (p. 204); however, an unattended demonstrative could help to clarify the problem without directly attributing it to someone, an important consideration when proposing a solution to that problem and expecting compliance from the offenders. The student’s use of this in (j) showed that unattended demonstratives mark a way for writers to deliver bad news with clarity but without assigning blame to an individual. Similarly, the student writer’s use of unattended this in (k) combined with the sentence’s sentiment (an apology) suggested a sophisticated means of attempting goodwill with an unhappy customer.

Finally, we found that social variables like gender and academic classification determined how students use (un)attended this in their technical writing. Our female students’ preference to attend this significantly more than the male students points toward the observation that females intuitively formalize their writing (Finegan & Biber, 2001). Likewise, our nonnative students’ preference to use attended this compared with their native speaker counterparts might relate to the prescriptive, formal English instruction the students received. The junior-level students consistently contrasted with the freshmen/sophomores and the seniors in our sample. Juniors overwhelming preferred to attend this in the initial and medial position, regardless of their gender or whether the text was an original or a revision. The consistency of these particular results is clear; however, we could only speculate at what they might mean. Further research on how these social variables impact student technical writing is needed. Additional studies could measure how students use other variants in their technical writing, including complementizers and hedges, and how these variables might impact text length as well as the size and density of students’ vocabularies.

We acknowledge that our sample size is too small to make any definitive conclusions about the effectiveness of (un)attended this across specific technical texts. Additional studies that focus on a single text type or a group of text types (e.g., correspondence) could better address the contexts in which writers commonly use unattended this. We also caution against equating our findings in student technical writing to practitioner technical writing. Previous research has documented the differences in learning writing in the classroom versus the workplace (Blakeslee, 2001; Freedman & Adam, 1996, 2000). A primary goal of classroom writing instruction is to scaffold knowledge; as soon as texts are evaluated and returned, students often advance to a new concept. On the other hand, workplace texts commonly evolve through extensive collaboration and undergo multiple revisions (Freedman & Adam, 2000). Results from corpus-linguistic studies have also noted differences between developing and expert (or published) writers; developing writers often make rhetorical choices that differ from the conventions adopted by experts (Römer, 2009b). Although our results do not provide insight into how practicing (or expert) technical writers apply (un)attended this, we believe they do provide teachers and tutors with some understanding of how their students think about writing. The content devoted to teaching demonstratives in technical communication textbooks is prescriptive and often negatively implicates writers who do not follow those rules. Our results demonstrate that students are engaged in a thought process when deciding whether to follow a demonstrative with a noun phrase. Understanding this process is a fundamental first step in developing (and then testing and revising) new technical writing pedagogies.

In response to our study, we encourage textbook authors to reconsider their discussion of (un)attended this and that instructors broaden how they teach this variant to students. We also acknowledge that editorial demands influence this call for enhanced instruction. Revising discussions of (un)attended this in current textbooks would require additional content on this variant as well as others that are likely in need of revisiting. Unfortunately, the length of these relevant discussions in popular technical communication textbooks appears to be significantly shortening rather than expanding. Wolfe, Britt, and Poe Alexander (2011) observed that the pages devoted to style, mechanics, and usage have remained stagnant in some technical communication textbooks and have been dramatically reduced or completely dropped in others (pp. 120–121). We noticed similar trends while working on our textbook analysis, and Woolever (2005) confirmed that the disappearance of the grammar section in her text’s most recent edition was related to an editorial suggestion (personal communication, August 6, 2011). Technical communication instructors must reconcile teaching grammar that reflects authentic context and use with these evolving editorial policies.

We hope our investigation of (un)attended this in technical and academic writing provides instructors, tutors, and students with materials for enhanced instruction and learning, including a list of verbs most significantly associated with the variant (see Figure 1), insights into where this is commonly placed within the sentence, and the rhetorical situations that might benefit from an unattended demonstrative. Further, Mackiewicz (1999) classified prescription-breaking discussions as those that eschewed traditional usage rules (p. 57). We identified none of these entries on (un)attended this in technical communication textbooks, so we conclude with a working discussion based on the findings as well as other empirical investigations of this variant:

The decision to follow the demonstrative pronoun this with a noun phrase can impact the clarity and acceptance of your message. Pronoun use can help you avoid excess repetition of nouns, but pronouns with no clear antecedent can impair how readers interpret your intended meaning.

The unattended this occurs frequently in both technical and academic writing and usually in the sentence-initial position (e.g., This is an example of sentence-initial position). The verbs “be” and “may” most often accompany the unattended this in academic writing; however, a variety of other verbs accompany this in technical writing, including “cause,” “allow,” and “help.”

The primary function of an unattended this is to summarize information that you have previously discussed. For example, “We appreciate your thorough plan for managing the construction of our hotel chain. This will help us make a decision about which consulting team to hire.”

Also, when referring to a situation that you believe readers might react negatively to, an unattended this can function as a buffer that communicates the targeted message without implicating anyone for causing the problem. For example, “The photocopy budget is deficient by several hundred dollars as a result of personal copy use. This will result in a change to copy-making policy.”

Finally, an unattended this can signal a justification, a statement that makes explicit the benefit of compliance. You often need to justify information that requires readers to accept or implement change, such as a new work policy or revising a feasibility report. For example, “All employees will be assigned a personal 3-digit PIN. This will help you monitor your photo-making use and preserve funds that the company can apply to its employees in other, more beneficial ways.”

Remember that a misuse of (un)attended this could muddle the message you are trying to convey. Always be aware of how and why you’re using unattended demonstratives.

Similar to Jordan’s (1999) claim, we believe the current, prescriptive discussions of (un)attended this provide technical communicators with a foundation for effective instruction; however, this instruction must also emphasize variations related to purpose and situation. This will help students become more mindful of how grammar and style shape their message.

ACKNOWLEDGMENTS

The authors would like to thank the TCQ reviewers and editorial staff for their comments on earlier drafts. Thanks to Cheryl Harrell and Carrie Klein for their help with coding the sample. Finally, we appreciate the University of North Texas’s support of this research through a Research Initiative Grant.

NOTES

1. References to (un)attended this relate to both the attended and unattended options of the variant. References to unattended this only relate to instances in which the demonstrative is left unattended.
2. The report assignment asked students to consider how they would revise their procedures for a specific multicultural audience. We chose to classify this text type as “academic” because the focus of the assignment was to learn, synthesize, and apply course material. Carter, Ferzli, and Wiebe (2004) made a similar classificatory decision regarding the biology laboratory report.
3. Technically speaking, any association measure, including the most widely known chi-squared test, can be employed. Here, we follow Stefanowitsch and Gries’s (2003, pp. 217–218) suggestion to use the Fisher-Yates exact test.
4. We computed the logistic regression in R using the function glm to obtain logistic models, the function ANOVA (model.glm, type = “III”, test.statistic = “Wald”) for model comparisons, and lrm(formula = formula(model.glm), x = T, y = T, linear.predictors = T) to obtain a summary of the predictive power of each predictor and the model as a whole.
5. For ease of exposition, Table 5 contains only significant predictor levels. Appendix D gives a complete overview of the results, including all predictor levels, standard errors, Wald’s z, and confidence intervals.

REFERENCES

1. Adamson, H. D. (1992). Social and processing constraints on relative clauses. American Speech, 67(2), 123–133.
2. Anderson, P. V. (2010). Technical communication: A reader-centered approach (7th ed.). Boston, MA: Wadsworth.
3. Anthony, L. (2006). Developing a freeware, multiplatform corpus analysis toolkit for the technical writing classroom. IEEE Transactions on Professional Communication, 49(3), 275–286.
4. Anthony, L. (2011). AntConc (Version 3.2.2). Tokyo, Japan: Waseda University. Retrieved from http://www.antlab.sci.waseda.ac.jp
5. Barker, T., & Matveeva, N. (2006). Teaching intercultural communication in a technical writing service course: Real instructors’ practices and suggestions for textbook selection. Technical Communication Quarterly, 15, 191–214. doi:10.1207/s15427625tcq1502_4
6. Blakeslee, A. M. (2001). Bridging the workplace and the academy: Teaching professional genre through classroom-workplace collaborations. Technical Communication Quarterly, 10, 169–192. doi:10.1207/s15427625tcq1002_4
7. Carter, M., Ferzli, M., & Wiebe, E. (2004). Teaching genre to English first-language adults: A study of the laboratory report. Research in the Teaching of English, 38, 395–419.
8. Connatser, B. R. (2004). Reconsidering some prescriptive rules of grammar and composition. Technical Communication, 51, 264–275.
9. Conrad, S. (2002). Corpus linguistic approaches for discourse analysis. Annual Review of Applied Linguistics, 22, 75–95. doi:10.0.3.249/S0267190502000041
10. Finegan, E., & Biber, D. (2001). Register variation and social dialect variation: The register axiom. In P. Eckert & J. R. Rickford (Eds.), Style and sociolinguistic variation (pp. 235–267). Cambridge, UK: Cambridge University Press.
11. Firth, J. R. (1957). Papers in linguistics, 1934–1951. London, UK: Oxford University Press.
12. Freedman, A., & Adam, C. (1996). Learning to write professionally: “Situated learning” and the transition from university to professional discourse. Journal of Business and Technical Communication, 10, 395–427. doi:10.1177/1050651996010004001
13. Freedman, A., & Adam, C. (2000). Write where you are: Situating learning to write in university and workplace settings. In P. Dias & A. Paré (Eds.), Transitions: Writing in academic and workplace settings (pp. 31–60). Cresskill, NJ: Hampton Press.
14. Fries, C. C. (1940). American English grammar. New York, NY: Appleton-Century.
15. Geisler, C., Kaufer, D. S., & Steinberg, E. R. (1985). The unattended anaphoric “this”: When should writers use it? Written Communication, 2(2), 125–155. doi:10.1177/0741088385002002002
16. Goldberg, A. E. (1995). Constructions: A construction grammar approach to argument structure. Chicago, IL: University of Chicago Press.
17. Goldberg, A. E. (2006). Constructions at work: The nature of generalization in language. Oxford, U.K.: Oxford University Press.
18. Gries, S. Th. (2007). Coll.analysis 3.2. A program for R for Windows 2.x.
19. Gries, S. Th., & Stefanowitsch, A. (2004). Extending collostructional analysis: A corpus-based perspective on ‘alternations.’ International Journal of Corpus Linguistics, 9(1), 97–129.
20. Harrell, F. E. (2011). Regression modeling strategies: With applications to linear models, logistic regression, and survival analysis. New York, NY: Springer.
21. Jaeger, T. F. (2010). Redundancy and reduction: Speakers manage syntactic information density. Cognitive Psychology, 61(1), 23–62. doi:10.1016/j.cogpsych.2010.02.002
22. Johnson-Sheehan, R. (2012). Technical communication today (4th ed.). New York, NY: Longman.
23. Jordan, M. P. (1999). “Unattached” clauses in technical writing. Journal of Technical Writing and Communication, 29, 65–93. doi:10.2190/41PB-WPVV-0VXY-JM1Q
24. Kolin, P. C. (2010). Successful writing at work (9th ed.). Boston, MA: Wadsworth.
25. Lannon, J. M. (2009). Technical Communication (11th ed.). New York, NY: Longman.
26. Mackiewicz, J. (1999). Gauging prescriptivism in writing handbooks. IEEE Transactions on Professional Communication, 42(1), 55–58.
27. Markel, M. (2012). Technical communication (10th ed.). New York, NY: Bedford/St. Martin’s.
28. McKenna, B., & Thomas, G. (1997). A survey of recent technical writing textbooks. Journal of Technical Writing and Communication, 27, 441–452. doi:10.2190/CGA9-CVJY-82CX-AEFJ
29. Miller, C. R. (1992). Textbooks in focus: Technical writing. College Composition and Communication, 43(1), 111–117.
30. Moskovit, L. (1983). When is broad reference clear? College Composition and Communication, 34(4), 454–469.
31. Oliu, W. E., Brusaw, C. T., & Alred, G. J. (2010). Writing that works: Communicating effectively on the job. Boston, MA: Bedford/St. Martin’s.
32. Orr, T. (2006). Introduction to the special issue: Insights from corpus linguistics for professional communication. IEEE Transactions on Professional Communication, 49, 213–216.
33. Pfeiffer, W. S., & Adkins, K. E. (2010). Technical communication: A practical approach (7th ed.). Upper Saddle River, NJ: Longman.
34. Raman, M., & Sharma, S. (2005). Technical communication: Principles and practice (3rd ed.). Oxford, U.K.: Oxford University Press India.
35. Reep, D. C. (2010). Technical writing: Principles, strategies, and readings (8th ed.). Boston, MA: Longman.
36. Riordan, D. G. (2005). Technical report writing today (9th ed.). New York, NY: Houghton-Mifflin.
37. Römer, U. (2009a). English in academia: Does nativeness matter? Anglistik: International Journal of English Studies, 20(2), 89–100.
38. Römer, U. (2009b). The inseparability of lexis and grammar: Corpus linguistic perspectives. Annual Review of Cognitive Linguistics, 7(1), 141–163.
39. Römer, U., & Wulff, S. (2010). Applying corpus methods to writing research: Explorations of MICUSP. Journal of Writing Research, 2(2), 99–127. Retrieved from http://www.jowr.org/articles/vol2_2/JoWR_2010_vol2_nr2_Roemer_Wulff.pdf
40. Stefanowitsch, A., & Gries, S. Th. (2003). Collostructions: Investigating the interaction between words and constructions. International Journal of Corpus Linguistics, 8(2), 209–243.
41. Swales, J. M. (2005). Attended and unattended “this” in academic writing: A long and unfinished story. ESP Malaysia, 11, 1–15.
42. Wolfe, J. (2009). How technical communication textbooks fail engineering students. Technical Communication Quarterly, 18, 351–375. doi:10.1080/10572250903149662
43. Wolfe, J., Britt, C., & Poe Alexander, K. (2011). Teaching the IMRaD genre: Sentence combining and pattern practice revisited. Journal of Business and Technical Communication, 25, 119–158. doi:10.1177/1050651910385785
44. Woolever, K. R. (2005). Writing for the technical professions (3rd ed.). New York, NY: Longman.
45. Wulff, S., & Gries, S. Th. (2011). Corpus-driven methods for assessing accuracy in learner production. In P. Robinson (Ed.), Second language task complexity: Researching the cognition hypothesis of language learning and performance. Amsterdam, The Netherlands: John Benjamins. Retrieved from http://www.linguistics.ucsb.edu/faculty/stgries/research/2011_SW-STG_AccuracyLearnerProd_L2TaskComplexity.pdf
46. Wulff, S., & Römer, U. (2009). Becoming a proficient academic writer: Shifting lexical preferences in the use of the progressive. Corpora, 4(2), 115–133. doi:10.3366/E1749503209000276
47. Wulff, S., Römer, U., & Swales, J. (2012). Attended/unattended this in academic student writing: Quantitative and qualitative perspectives. Corpus Linguistics and Linguistic Theory, 8(1), 129–157. doi:10.1515/cllt-2012-0006

APPENDIX A. TECHNICAL COMMUNICATION TEXTBOOKS CONSULTED FOR OUR REVIEW OF (UN)ATTENDED THIS DISCUSSIONS

Author and Year of Publication	Title (Edition)	Discussion Page Number
Anderson, 2010	Technical communication: A reader centered approach (7th ed.)	226
Burnett, 2005	Technical communication (6th ed.)	None found
Gerson & Gerson, 2011	Technical writing: Process and product (7th ed.)	None found
Houp, Pearsall, Tebaux, & Dragga, 2006	Reporting technical information (11th ed.)	None found
Johnson-Sheehan, 2012	Technical communication today (4th ed.)	A–8
Kolin, 2010	Successful writing at work (9th ed.)	704–705
Lannon, 2009	Technical communication (11th ed.)	217
Markel, 2012	Technical communication (10th ed.)	219
Oliu, Brushaw, & Alred, 2010	Writing that works: Communicating effectively on the job (10th ed.)	629–630, 671
Pfeiffer & Adkins, 2010	Technical communication: A practical approach (7th ed.)	646–647
Raman & Sharma, 2005	Technical communication: Principles and practice (3rd ed.)	204
Reep, 2010	Technical writing: Principles, strategies, and readings (8th ed.)	497
Riordan, 2005	Technical report writing today (9th ed.)	559
Woolever, 2005	Writing for the technical professions (3rd ed.a)	470

^aWe used Writing for the Technical Professions, 3rd edition, because of its thorough discussion of (un)attended this. This discussion, along with the discussions of other ambiguous pronouns, is not featured in the more recent 4th edition.

APPENDIX B. NUMBER OF TEXTS IN THE DATA SAMPLE BY STUDENT BACKGROUND, TEXT CLASSIFICATION, AND DRAFT

Student	Gender	Year	L1/L2	Type of Majora	Technical	Academic	Original	Revision	Total
A	f	senior	L1	STEM	12	4	9	7	16
B	f	senior	L1	nonSTEM	12	6	11	7	18
C	f	junior	L1	nonSTEM	12	7	12	7	19
D	m	senior	L1	STEM	12	6	11	7	18
E	f	senior	L1	nonSTEM	11	4	8	7	15
F	f	senior	L1	nonSTEM	12	6	11	7	18
G	m	senior	L1	nonSTEM	6	2	1	7	8
H	f	junior	L1	nonSTEM	11	7	12	6	18
I	m	senior	L2	STEM	11	5	10	6	16
J	f	freshman	L1	STEM	12	6	11	7	18
K	m	junior	L1	STEM	6	4	5	5	10
L	f	sophomore	L1	STEM	12	5	10	7	17
M	f	senior	L1	STEM	11	7	11	7	18
N	f	senior	L1	nonSTEM	12	6	11	7	18
O	f	senior	L1	nonSTEM	11	6	11	6	17
P	f	junior	L1	STEM	6	4	10	0	10
Q	f	junior	L1	nonSTEM	12	6	11	7	18
R	m	freshman	L1	STEM	12	6	11	7	18
Group	f/m	Mixed	Mixed	Mixed	10	0	2	8	10
Total					203	97	178	122	300

^aStudents majoring in agricultural and applied economics; anthropology; speech, language, and hearing sciences; psychology; chemistry; and political science were classified as STEM majors. Students majoring in special education, human development and family studies, retailing, personal financial planning, and interior design were classified as nonSTEM majors.

APPENDIX C. VERBS DISTINCTIVELY ASSOCIATED WITH ATTENDED AND UNATTENDED THIS

Attended This Verb Lemma	plog
learn (41:0)	7.333
have (55:6)	4.831
complete (26:0)	4.628
use (38:7)	2.245
write (24:3)	2.112
encounter (16:1)	2.034
meet (17:2)	1.636
work (16:2)	1.501
hope (8:0)	1.416

Unattended This Verb Lemma	plog
mean (0:27)	−13.008
be (317:261)	−11.769
cause (4:18)	−5.417
allow (13:27)	−5.048
help (21:29)	−3.581
give (12:19)	−2.904
indicate (0:6)	−2.863
think (6:10)	−1.799
create (4:8)	−1.727
encourage (0:3)	−1.429
identify (0:3)	−1.429
do (33:27)	−1.414
imply (2:5)	−1.342
bring (1:4)	−1.342

Note. The numbers in parentheses are the observed frequencies of the verb lemma with attended and unattended this, respectively.

APPENDIX D. COMPLETE OUTPUT OF THE MINIMAL ADEQUATE LOGISTIC REGRESSION MODEL

Predictor	Coefficient	Standard Error	Wald z	p	2.5% CI	97.5% CI
Position = final	−2.097	0.468	−4.480	0.000	−3.112	−1.248
Position = medial	−1.313	0.325	−4.050	0.000	−1.976	−0.698
Text Classification = technical × Draft = revision	−1.111	0.237	−4.690	0.000	−1.576	−0.647
Gender = male × Year = senior	−0.898	0.321	−2.800	0.005	−1.530	−0.270
Year = junior	−0.867	0.281	−3.090	0.002	−1.423	−0.321
Gender = male × Year = junior	−0.797	0.412	−1.930	0.053	−1.614	0.002
Language = nonnative	−0.767	0.323	−2.370	0.018	−1.420	−0.149
Draft = revision × Year = senior	−0.628	0.294	−2.130	0.033	−1.207	−0.052
Position = final × Year = senior	−0.289	0.590	−0.490	0.624	−1.434	0.915
Position = final × Year = junior	−0.199	0.875	−0.230	0.820	−2.200	1.399
Distinctiveness	−0.088	0.009	−9.660	0.000	−0.106	−0.070
Draft = revision × Year = junior	−0.083	0.369	−0.220	0.822	−0.808	0.638
…
Position = medial × Year = junior	1.461	0.417	3.500	0.001	0.655	2.295
Draft = revision	1.168	0.282	4.150	0.000	0.619	1.724
Gender = male	0.678	0.273	2.490	0.013	0.145	1.216
Position = medial × Year = senior	0.556	0.354	1.570	0.117	−0.121	1.272
Year = senior	0.432	0.216	2.010	0.045	0.011	0.857
Text Classification = technical	0.197	0.152	1.290	0.197	−0.104	0.494

---

Ryan K. Boettger is an assistant professor in the Department of Linguistics and Technical Communication at the University of North Texas.

Stefanie Wulff is an assistant professor in the Linguistics Department at the University of Florida. Boettger and Wulff are the creators of TechCorp, a soon-to-be publicly released corpus of student technical writing.