2.2.1 Some Core Issues in Linguistic Terminology

The use of linguistic terminology across subdisciplines is influenced by a range of factors, as noted by Bugarski (Reference Bugarski1983), and two of the most crucial ones in this regard are the linguistic paradigm as well as the individual definition applied in a study. A core conflict that results from the inconsistency of how terms are used is the ‘distinction between standardisation and unification, allowing for the coexistence of two or more internally unified and partly overlapping standards’ (Bugarski Reference Bugarski1983: 69).Footnote ¹ While an outcome of terms emerging in specific paradigms may be a ‘typically short lifespan of the terms coined by the proponents of particular theoretical frameworks’ (Trask Reference Trask1993: viii), the situation is probably more complex when semantic ambiguity (as opposed to neologism) is involved.Footnote ²

Another important issue is the applicability of terms across disciplines, languages, time periods, etc. As Vermeer (Reference Vermeer1971: 14–15) notes, definitions are necessarily products of a given setup of contextual conditions, including space and time. He suggests table as an example: how can we define table without missing some members of this category? His publication on the topic slightly predates prototype theory (Rosch Reference Rosch1973) but is strongly reminiscent of the idea that fuzziness is often inevitable.

While terms may be coined to describe a particular phenomenon, already existing terms may also be redefined (either intentionally or as the result of semantic shift), a practice that ‘may or may not have an effect on how the term is applied in practice so that the outcome is either polysemy or inconsistency’ (Mugdan Reference Mugdan, Niederehe and Koerner1990: 57). A recent example of intentional change is the reframing of certain concepts in Rüdiger (Reference Rüdiger2019), who suggests using the qualifiers ‘plus‑’ and ‘minus‑’ instead of terms such as ‘overuse’ or ‘underuse’ for specific linguistic features (see Rüdiger et al. Reference Rüdiger, Leimgruber and Tseng2022 for an example of this terminology in use). While achieving ‘true’ terminological neutrality is probably impossible due to the individual and often purpose-driven definition of linguistic terms, reducing their evaluative or even prescriptive connotations is in line with a move towards descriptivism. This is especially relevant in the context of linguistic disciplines – such as World Englishes and sociolinguistics – that deal with varieties that historically have been or could be subject to discrimination. As this brief summary has shown, linguistic terminology and the encoded knowledge systems are rather complex and the factors influencing them are manifold. In order to gain insight into how the terms central to this edited volume are used, we next provide a brief overview of how the terms are framed in selected dictionaries and grammars before introducing our study.

2.2.2 ‘Canonical’ and ‘Non-Canonical’ in Selected Dictionaries and Grammars

Some of the issues introduced above also apply to the ‘canonical’ and ‘non-canonical’ pair of terms. An important concept strongly linked to language ideology is that of the ‘native speaker’, as noted by Hackert (Reference Hackert2012: 30): ‘at least according to a number of authors, what we are looking at if we look at the English native speaker is an imaginary or political construct, something which is discursively constituted and created, and which is about attitudes, affiliation, and social identity rather than about linguistic competence’. ‘Canonical’ and ‘non-canonical’ also fall into the category of terms that may carry intended or unintended ideological weight, since they may be consciously or subconsciously linked to ideas of correctness, producing ‘native-like’ language, following specific rules, and so forth; Hackert (Reference Hackert2012: 30) links the native speaker ideology both to the history of Western linguistics and to other ideologies, including standard ideology. A look into the entry for ‘canonical’ in the online version of the Oxford English Dictionary (OED) reveals the following definition as the most (and only) relevant one for the syntactic context:Footnote ³

The question to what extent ‘canonical’ and ‘non-canonical’ suggest or even imply superiority of one syntactic variant over another obviously depends on how the terms are defined and applied in any given study, but their morphological composition (which includes reference to the notion of the ‘syntactic canon’) inevitably points to interpretations in the sense of ‘according to the canon’ or ‘not according to the canon’. It is interesting to note, then, that some definitions of related terms include ‘canonical’ but not ‘non-canonical’, a case in point being Hickey’s entry for ‘word order’ in A dictionary of varieties of English:

While it would be fascinating to consider how further dictionaries and grammars treat the semantic field of canonicity, a comprehensive overview goes far beyond the scope of this chapter. However, we consider it crucial to address at least briefly how the three major grammars of English (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999;Footnote ⁴ Huddleston & Pullum Reference Huddleston and Pullum2002) deal with ‘canonical’, ‘non-canonical’, and related terms in the context of word order variation. As references with citations in the tens of thousands,Footnote ⁵ all three titles have had a significant impact not only on the dissemination but also on the stabilisation of linguistic terminology across subdisciplines. The first of the three mentions ‘canonical’ early in the book:

The reasoning behind equating the ‘simple declarative sentence’ with the ‘canonical form’ remains unclear, however. The term ‘non-canonical’ is not used in Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985) at all. Similarly, in their introduction to word order, Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999) use the terms ‘unusual’ and ‘marked’ but do not mention ‘canonical’, ‘non-canonical’, or ‘canonicity’; the ‘typical’ form of a sentence is addressed with reference to the fixed nature of English word order:

Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999: 896) note that ‘marked’ word order may achieve effects that what we call ‘canonical’ syntactic patterns cannot; such effects include achieving emphasis, contrast, and structuring the information flow.

Finally, in the relevant chapter of the Cambridge grammar (Huddleston & Pullum Reference Huddleston and Pullum2002), Ward et al. (Reference Ward, Birner, Huddleston, Huddleston and Pullum2002) contrast ‘canonical’ and ‘information-packaging’ constructions:

It is interesting to consider how Ward et al. (Reference Ward, Birner, Huddleston, Huddleston and Pullum2002) contextualise syntactic (non-)canonicity. By pointing out that ‘the syntax makes available different ways of “saying the same thing”, with the various versions differing in the way the content is organised informationally’ (Ward et al. Reference Ward, Birner, Huddleston, Huddleston and Pullum2002: 1365; emphasis added), they deliberately link their understanding of (non-)canonicity to syntactic variation and in turn to both Labov (Reference Labov1972) and, accordingly, sociolinguistic variation as well as register variation. This framing overtly suggests understanding different syntactic alternatives as variants that may be influenced by intra- and extra-linguistic factors.

As mentioned earlier, it is not possible to trace the entire history of ‘canonical’, ‘non-canonical’, ‘canonicity’, etc. in this section, nor was that the goal. However, our spot check of relevant publications has revealed certain tendencies: (i) ‘canonical’ appears to be favoured in grammars as a term over ‘non-canonical’, sometimes to the extent of ‘non-canonical’ not being named as the counterpart; (ii) ‘canonical’ is often framed as ‘basic’, ‘essential’, or ‘close to the standard’; (iii) canonicity is frequently associated with unmarkedness, the standard, and syntactic variation. These observations represent important reference points in our empirical analysis below.

2.3.1 Data and Method

For the case study, we collected all articles published from 2012 to 2021 in the journals Corpora, the Journal of English Linguistics, the Journal of Germanic Linguistics, the Journal of Historical Linguistics, the Journal of Linguistics, and Syntax. An overview of the journals, including the publisher as well as the number of articles contained in our corpus, is provided in Table 2.1.

The selection of the journals to be included was driven by four main factors: (a) the relevance of the journals to the edited collection, (b) the relevance of the journals to the study of syntax more widely, (c) the accessibility of the journals to the authors, and (d) the aim to include research on historical varieties of English as well as languages other than English. Corpora is, as the name suggests, devoted to studies in corpus linguistics and, accordingly, has a strong empirical and methodological focus. The Journal of Germanic Linguistics, the Journal of Linguistics, and Syntax share an interest in typological research and a tendency to include contributions from generative grammar and related approaches, although none of the journals is exclusive in this regard. While the former two journals certainly feature a substantial number of contributions dealing with syntactic phenomena, they also include, among others, articles in morphology and phonetics. The Journal of English Linguistics is thematically the broadest journal in our corpus and covers topics as diverse as historical linguistics, phonetics, World Englishes, and, of course, syntax. In contrast to the three journals just mentioned, the focus of contributions to the Journal of English Linguistics is generally rather usage-based and empirical. Finally, the Journal of Historical Linguistics invites papers on all facets of historical linguistics across languages, although the website description highlights that ‘contributions in areas such as diachronic corpus linguistics or diachronic typology are … particularly welcome’.Footnote ⁶

An issue in compiling the corpus has been the question of what to include and what to exclude, since the journals also publish text types other than research articles that may or may not be relevant for our study. We decided to include all research articles, introductions to special issues, and remarks/short notes and similar smaller publications, but to exclude book reviews, editorial notes, and annotated bibliographies (published separately, for instance as the final part of special issues). This resulted in a corpus of 783 publications, with an average of 130.5 publications per journal (SD = 30.98). The articles were originally available in PDF format and were then processed via OCR to create txt-files suitable for corpus analysis. Random inspection of the corpus files created in this way revealed some errors in the text files which were due to the automatic processing of the data (e.g., OCR artefacts). Overall, however, these problems seemed to be only marginal in nature, and we decided to forgo manual correction in light of feasibility. In addition, the automatised conversion process included page numbers and page headers and we therefore do not present word counts for the subcorpora and the corpus overall but instead use articles as basis for normalisation. Even though article length might vary, we believe this provides a reasonable basis for comparison for the purposes of the current study.

For the analysis, we extracted all instances of the words listed in Table 2.2 using AntConc. The first three groupings consist of oppositional pairs: ‘canonical’ and ‘non-canonical’, ‘standard’ and ‘non-standard’, and ‘marked’ and ‘unmarked’.Footnote ⁷ The searches for ‘canonical’ and ‘non-canonical’ also included the adverb forms ‘canonically’ and ‘non-canonically’, but most of the uses recorded were either as attributive or as predicative adjectives. While ‘non-standard’ is typically used as an adjective as well, ‘standard’ may also be employed as a noun, which also explains the higher general frequency observed for this item in the analysis. As the data were not part-of-speech tagged, we present no further details on word class usage, but see Section 2.3.3 for a collocational analysis of these lexical items. The nouns ‘canonicity’ and ‘markedness’ reflect the semantic relation between ‘canonical’ and ‘marked’ as well as their negated counterparts; however, in most general terms, ‘canonicity’ refers to the ‘typical’, whereas ‘markedness’ refers to the ‘atypical’.

Table 2.2Terms considered in the analysis

Table 2.2 Long description

The table is divided into two columns with the labels: grouping and terms. There are 4 rows arranged from top to bottom with the corresponding data. Their details are listed as follows:

• For pair 1, the terms are canonical and non-canonical

• For pair 2, the terms are marked and unmarked

• For pair 3, the terms are standard and non-standard

• For additional terms, the terms are canonicity, markedness, and syntactic variation.

^a The term ‘uncanonical’ occurred only once in the dataset and was thus not considered for further analysis.

^b The term ‘non(-)marked’ occurred only three times in the dataset and was thus not considered for further analysis.

Of course, further terms and constructions, such as ‘alternative / basic / different / infrequent / (a)typical / uncommon / (un)usual word order’, are potentially used (near-)synonymously with the lexical items investigated in our analysis. For reasons of space, we concentrate here on the target items as specified above and as listed in Table 2.2, but further research on terminological choices should account for these alternatives as well as the actual terminological definitions given in the papers.

It needs to be pointed out that our analysis is based on the articles in their entirety, that is, including the bibliography (but also abstract, bionote, etc.). At times, our lexical target items do indeed occur in the reference section of an article, for example, when an article cites Jenny Cheshire’s Reference Cheshire1987 Linguistics article with the title ‘Syntactic variation, the linguistic variable, and sociolinguistic theory’ (emphasis added). As the bibliography is an essential part of academic manuscripts and referencing specific linguistic works evokes the words used in their titles and thus plays a role in the reinforcement of linguistic terminology, we decided to not exclude the bibliography sections from the corpus.

Last but not least, as will also become clear in the analysis section, most of the terms, particularly those in pairs 1, 2, and 3, are also in usage outside the realm of syntax – and can, for example, also be found in the description of semantic phenomena. In addition, ‘marked’ and ‘standard’ (and to a certain degree ‘(non‑)canonical’) have further non-specialist meanings (e.g., marked used as a past tense verb form in the sentence ‘The presentation of the Dynamic Model in 2007 marked a major milestone in the rapidly emerging field of World Englishes’, JEngL2015)Footnote ⁸ or are used in fixed specialised constructions which are unrelated to specific fields (e.g., ‘standard deviation’). While such uses affect the results to a certain extent, the majority of them pertain to the present study. Thus, manually weeding out irrelevant hits (an extremely time-consuming task) was not considered beneficial. We will expand on these aspects further in Section 2.3.3 on the qualitative analysis of our data.

2.3.2 Quantitative Analysis

First, we consider the frequencies of the terms listed in Table 2.2. Figure 2.1 shows the normalised frequencies of the first two pairs across all six journals.Footnote ⁹ We normalised the figures per 100 articles instead of per a certain word count (see Section 2.3.1); the values for all terms as well as their range (i.e., occurrence across articles in each journal) are listed in Appendix A (Table 2.5).

Figure 2.1

Term frequency per 100 articles per journal for ‘canonical’/‘non-canonical’ and ‘marked’/‘unmarked’

Figure 2.1 Long description

The six journals are represented using different symbols:

• Cor represented by grey circles

• JEngL represented by dark triangles

• J G L represented by black squares

• J H L represented by light crosses

• JoL represented by a hollow box

• Syn represented by a light star

The horizontal axis displays four term categories: canonical, non-canonical, marked, and unmarked. The vertical axis shows token counts, ranging from 0 to 500 in increments of 100. The graph is divided into two sections with a group of lines marked on the left and another set of lines marked on the right. The data from the graph, from left to right, is as follows:

• The token values for JoL, J G L, Syn, JEngL, J H L, and Cor are 350, 150, 100, 90, 90, and 20 for the canonical term. The corresponding token values for the non-canonical term are 100 and a value that ranges between 50 and 0.

• The token values for JoL, J H L, Syn, J G L, JEngL, and Cor are 460, 420, 350, 250, 250, and 250 for the marked term. The corresponding token values for the unmarked term are 200, 160, 150, 100, and a value that ranges between 90 and 70.

Figure 2.1 shows that, for both term pairs, the non-negated term occurs more frequently than the negated term. Comparing ‘canonical’ and ‘non-canonical’ across all six journals using a chi-squared test reveals statistically highly significant differences (χ² = 24.074, df = 4, p < 0.005, Cramer’s V = 0.130389). However, with the exception of the Journal of Linguistics, the differences in frequency between ‘canonical’ and ‘non-canonical’ appear less extreme than the differences between ‘marked’ and ‘unmarked’. For this pair, the statistical difference across all journals is also significant, albeit to a lesser degree when compared to ‘canonical’ and ‘non-canonical’; the effect size is noticeably small for the frequency difference (χ² = 10.755, df = 4, p < 0.05, Cramer’s V = 0.05684054).

The higher frequency of the non-negated term appears to be even more extreme in the ‘standard’/‘non-standard’ pair, which is depicted in Figure 2.2 and for which the differences across the journals are, again, statistically highly significant (χ² = 179.75, df = 4, p < 0.005, Cramer’s V = 0.2176041).

Figure 2.2

Term frequency per 100 articles per journal for ‘standard’/‘non-standard’

Figure 2.2 Long description

The six journals are represented using different symbols:

• ● Cor represented by grey circles

• ● JEngL represented by dark triangles

• ● J G L represented by black squares

• ● J H L represented by light crosses

• ● JoL represented by a hollow box

• ● Syn represented by a light star

The horizontal axis displays two, standard and non-standard. The vertical axis shows token counts, ranging from 0 to 1.1K in increments of 100. A series of lines is marked at the center. The data from the graph, from top to bottom, is as follows:

• The token values for J G L, JEngL, JoL, Syn, J H L, and Cor are 1050, 600, 500, 350, 260, and 200 for the standard term. The corresponding token values for the non-standard terms are 100, 150, and a value that ranges between 50 and 0.

In Corpora, 28.24 tokens of ‘canonical’ and 12.21 tokens of ‘non-canonical’ occur per 100 articles, which means that the non-negated form occurs slightly more than twice as often. In contrast, ‘standard’ occurs at 204.58 and ‘non-standard’ at 50.38 tokens per 100 articles, meaning that the non-negated form occurs four times as often. However, as mentioned above, this is likely due to the multiple word class membership of ‘standard’.

While the tendency of the non-negated form being more frequent than the negated form appears consistent throughout, it is important to consider dispersion as an additional measure. When authors decide on a set of terms in their study, it seems likely that they stick to these terms, which means that investigating dispersion may reveal potential imbalances in term usage across journals. To illustrate this phenomenon, Figure 2.3 shows an X-ray plot of the dispersion of ‘canonical’, ‘non-canonical’, ‘marked’, and ‘unmarked’ in the articles published in Syntax in 2021.

Figure 2.3

X-ray plot showing the dispersion of ‘canonical’, ‘non-canonical’, ‘marked’, and ‘unmarked’ in articles published in Syntax in 2021

Figure 2.3 Long description

The x-ray plot is divided into four sections. The sections at the top are labeled as canonical, non-canonical, marked, and unmarked. The horizontal axis marks the relative token index. Each section at the bottom marks the relative token index ranging from 0 to 1, in increments of 0.25. The vertical axis on the right side is labeled with document names, including 1.txt, 10.txt, 11.txt, 12.txt, 13.txt, 3.txt, 5.txt, 6.txt, 14.txt, 7.txt, and 9.txt. Each row contains vertical lines that are marked, which indicate the relative position of each term in the particular document. The data from left to right in the document is as follows:

• Canonical are in 1.txt, 10.txt, 11.txt, 12.txt, and 13.txt.

• Non-canonical are in 12.txt and 6.txt.

• Marked are in all documents except 10.txt and 12.txt.

• Unmarked are in 1.txt, 11.txt, 13.txt, 7.txt, and 9.txt.

Overall, 14 publications are part of this corpus segment. However, as three publications do not contain any of the four items, only 11 files are included in the X-ray plot. Six of the 11 publications (54.55%) use a term from both pairs at least once; in some cases, a clear preference for one pair (e.g., ‘marked’/‘unmarked’ in file 13) is obvious. In addition to the X-ray plot, we calculated dispersion measures based on Gries (Reference Gries2008, Reference Gries, Paquot and Gries2020) and the corresponding R script. More precisely, we calculated the DP value (deviation of proportions) for the three pairs in a comparison of all six journals.Footnote ¹⁰ The advantages of DP over other dispersion measures are manifold and explained in detail in Gries (Reference Gries2008) and Gries (Reference Gries, Paquot and Gries2020); a key feature is that DP is able to deal with uneven corpus sizes, which makes it appropriate for our purposes.Footnote ¹¹ The values are shown below:

• ‘canonical’: DP = 0.2943403

• ‘non-canonical’ (incl. ‘noncanonical’): DP = 0.4083444

• ‘marked’: DP = 0.08008977

• ‘unmarked’: DP = 0.1218987

• ‘standard’: DP = 0.2161513

• ‘non-standard’ (incl. ‘nonstandard’): DP = 0.48796315

In general, lower DP values indicate a more even spread, and higher DP values indicate a more uneven spread across the dataset. While the low DP value for ‘marked’ is not surprising given its multi-word-class status, ‘unmarked’ has the second-lowest frequency, meaning that it is also comparatively evenly dispersed. ‘Non-canonical’ and ‘non-standard’ have the highest DP values, which, in general terms, means that they are the least evenly dispersed of the six terms. This suggests that their dispersion is ‘clumpy’, that is, they are not used evenly by a range of authors. ‘Canonical’ and ‘standard’ fall between the other terms mentioned so far.

Finally, the absolute and normalised frequencies per 100 articles of the terms ‘canonicity’, ‘markedness’, and ‘syntactic variation’ as well as their range across articles within each journal are presented in Table 2.3.

Table 2.3Absolute and normalised frequencies of ‘canonicity’, ‘markedness’, and ‘syntactic variation’ in six linguistic journals as well as range across articles within each journal

Table 2.3 Long description

The table is divided into 7 columns, each with the headers: terms, Cor or corpora, JEngl or Polish language, J G L or Dutch and German language, J H L or Finnish, JoL or Polish, Arabic, and Russian, and Syn or standard of analysis and comparison. It displays absolute frequencies A F, normalized frequencies N F, and occurrences with percentages for each term in each journal. The data from left to right in each row is as follows:

1. 1. For the term canonicity:

   • The corresponding values for A F are 8, 5, 7, 0, 11, and 0.

   • The corresponding values for N F are 6.11, 4.13, 7.22, 0, 5.64, and 0.

   • The corresponding values for occurrence are 2 in 131, 2 in 121, 3 in 97, 0 in 111, 8 in 195, and 0 in 128.

   • The corresponding values for percentage are 1.53, 1.65, 3.09, 0.00, 4.10, and 0.00.

2. 2. For the term markedness:

   • The corresponding values for A F are 53, 28, 34, 123, 399, and 41.

   • The corresponding values for N F are 40.46, 23.14, 35.05, 113.89, 204.62, and 32.03.

   • The corresponding values for occurrence are 5 in 131, 12 in 121, 16 in 97, 14 in 111, 55 in 195, and 24 in 128.

   • The corresponding values for percentage are 3.82, 9.92, 16.49, 12.61, 28.21, and 18.75.

3. 3. For the term syntactic variation:

   • The corresponding values for A F are 6, 43, 16, 12, and 9.

   • The corresponding values for N F are 4.58, 35.54, 6.49, 11.11, 9.23, and 7.03.

   • The corresponding values for occurrence are 4 in 131, 18 in 121, 8 in 97, 8 in 111, 16 in 195, and 10 in 128.

   • The corresponding values for percentage are 3.05, 14.88, 8.25, 7.21, 8.21, and 7.81.

Note. AF = absolute frequency; NF = normalised frequency; Occurrence = occurrence in number of articles.

The figures show that, in general, ‘canonicity’ is not a frequent term in any of the journals, with zero hits in the Journal of Historical Linguistics and Syntax. ‘Markedness’, on the other hand, occurs relatively frequently across the corpus, with the Journal of Linguistics and the Journal of Historical Linguistics boasting the highest frequencies. Finally, ‘syntactic variation’ is most frequently used in the Journal of English Linguistics, which underlines the journal’s tendency to favour usage-based, empirical, and often sociolinguistic contributions. However, it is important to be aware that individual preferences of authors, editorial guidelines, text type, etc. may all have an influence on how terms are used and defined; a closer look into the terms in use is provided in the next section.

2.3.3 Qualitative Analysis

In the following, we present the results of an n-gram analysis for the first three target pairs, grouped by the six journals and focusing on the top five bigrams (with the target item occurring in the first slot).Footnote ¹² The analysis was conducted using AntConc’s clusters/n-grams tool, with the minimum frequency set to two occurrences. The table with the full list of results can be found in Appendix B (Table 2.6). We use the results from the top five bigram analysis to inform our further analysis of specific constructions. It should be mentioned that absence from the top five bigram list does not automatically entail that the specific word combination is not used in a journal, as it could merely be less frequent than the top five listed. Where necessary, bigrams were extended to include following items (e.g., to complete noun phrases).

Pair 2: ‘Marked’ and ‘Unmarked’

According to our bigram analysis, ‘marked’ is predominantly used as a verb in our corpus and is usually followed by a preposition. The list of top five right collocates across the subcorpora only contains two non-prepositions: plural (n = 21; range = 1) in Corpora as well as and in the Journal of Linguistics (n = 52; range = 15). Across all subcorpora, ‘marked and’ occurs 125 times. Most frequent is the combination ‘marked and unmarked’ (n = 31; range = 10), setting up a similar contrast as seen above for ‘canonical and non-canonical’, see (3).

(3)

One of the most long-standing debates in the generative framework has hinged on the specification of the features and/or principles that motivate marked and unmarked syntactic orders.

(Syn2019)

In terms of prepositional right collocates, the preposition for is of interest to us here (occurring in all top five bigrams with ‘marked’ in the left position), as this potentially shows us what is being marked. In Table 2.4, we list all right collocates for ‘marked for’ and ‘unmarked for’ which occur at least three times across all subcorpora.

Table 2.4Right collocates of ‘marked for’ and ‘unmarked for’ occurring at least three times in the corpus

Table 2.4 Long description

The table is divided into two main columns and labeled marked for and unmarked for.

• Under marked for, several right collocates are listed along with their frequency: the, n = 33, tense, n = 15, case, n = 9, gender, number, past, n = 7, person, n = 6, progressive, n = 5, definiteness, feminine, n = 4, accusative, aspect, deletion, force, masculine, n = 3.

• Following these, the sub-column lists progressive, n = 5, absence, genitive, n = 4, dense, use, same, n = 2.

• Under unmarked for, several collocates are listed along with their frequency case, past, n = 4.

As can be seen in Table 2.4, and in contrast to example (3) above, ‘marked for’ clearly has the primary meaning of referring to the presence of morphological marking, that is, the authors refer to verbs, for example, being marked for the progressive, tense, or the past, and nouns and other parts of speech for case, gender, or definiteness. The same applies to ‘unmarked for’, which occurs four times, each followed by case or past. What is absent here is an underlying notion of typicality/frequency/standardness, which contrasts with how ‘canonical’ and ‘non-canonical’ are used in the corpus (see above).

The top five right collocates of ‘unmarked’ in five of the six journals surveyed contained form and/or forms. Here, ‘unmarked’ primarily seems to reference the absence of a particular kind of marking (usually one that would be expected). In (4), for example, an English verb used by a speaker is described as ‘unmarked’ for the past tense and in (5) the ‘unmarked form’ refers to a Korean noun which is used without a subject case marker (but which nonetheless is considered ‘the better option’).

(4)

She uses the unmarked form give to reference this past difficulty.

(JEngL2017)

(5)

Whereas both the case-marked and noncase-marked form are acceptable in (a), the unmarked form is the better option in (b).

(JoL2016)

This seems to contrast with the results for ‘un/marked for’ given above and indicates that more in-depth qualitative analysis is necessary to further disambiguate the different usages of these terms.

2.3.4 Meta-Discussion of Terminology

As can be seen in the use of constructions such as ‘I/we label’ (n = 24), ‘I/we term’ (n = 25), and ‘I/we call’ (n = 93),Footnote ¹⁵ article authors at times motivate their terminological choices or at least draw explicit attention to them. In a few cases, such as (6), this concerns the terms under consideration in this study. In (6), we find the author distinguishing between two agreement patterns, labelling one ‘non-standard’ and the other ‘uncommon’.

The author continues to set up a three-way contrast, between ‘standard’, ‘non-standard’, and ‘uncommon’ (see (7)). In this case, the terminological distinction is motivated by the ‘variable levels of exposure’ by speakers to the specific constructions.

Labelling one construction specifically as ‘uncommon’ might lead to the assumption that the other two are ‘common’ (albeit one of them is also non-standard) and it would be interesting to know why other potential terminological choices were rejected in this specific case (e.g., ‘canonical’/ ‘non-canonical’).

Finally, we reproduce in (8) an exceptionally long passage explaining the choice of an author to use the terms ‘canonical’ and ‘non-canonical’. Even though the author applies these terms to describe semantic phenomena, we find his reasoning of particular interest for our study as well. The terminological choice of ‘canonical’ and ‘non-canonical’ is motivated by the terms being perceived as (1) describing phenomena which are considered ‘more or less stable’, (2) less judgmental than the terms ‘good’ or ‘bad’, and (3) non-mutually exclusive (i.e., gradable). In addition, the terms are considered ‘necessarily imprecise’ – a quality considered both essential and problematic by the manuscript’s author.