The Grammatical Status of However

1. Introduction

The handy transitional word however seems to challenge writers more than the average conjunctive adverb. In between clauses, the term is the bane of English teachers and editors, who regularly cross out commas and insert semicolons before that pesky however. Popular grammar books and online grammar websites likewise admonish their readers not to use however as a coordinator, even though it sets up a contrast very similar to but. In such cases, these usage guides typically provide an unambiguous fix, based on revising punctuation. Consider (1) from GrammarErrors.com.

The very ubiquity and force of this prescription about how to correctly use however suggests that the rule is far from intuitive for many writers of standard edited English. And the “incorrect” sentence above indicates what writers are trying to do: they are granting however the ability to connect clauses with the power and the preceding comma that characterize a coordinator (although they often retain the comma after however). Bryan Garner, in his well-regarded Modern American usage (2016:472), labels this coordinating use of however “misuse,” and he captures the widely held belief that confusion about the standard use of however is generally relegated to “informal writing, student papers, and the like.”

However, relegating coordinating uses of however to “unschooled writing” or solely informal writing may overly minimize the phenomenon. ¹ Without a doubt, examples of coordinating or “misused” however are very difficult to find in formal writing that has been edited according to a house style guide (e.g., in newspapers, academic journals, etc.). But examples of however taking on a syntactically coordinating function (equivalent to but) and punctuated with a preceding comma pop up with notable regularity in, for example, informational pamphlets and official websites: in other words, sources that are relatively formal but may have been less formally edited. The Harvard Review, for instance, included on its submissions webpage in 2016 (although it has since been corrected) the formulation in (2).

The sentence in (3) appeared in the column “Ask Our Dietitians” in the print magazine Clean Eating (October 2015).

As we prepared to present some of this research at a conference, we received the confirmation letter from the hotel as given in (4).

That said, it is not impossible to find this coordinating however in well-respected newspapers like The New York Times. Consider (5) from the 2018 article “Teacher’s Gun Is Accidentally Fired During Public Safety Class, Injuring 3” (Caron 2018):

In sum, there is a clear, unambiguous usage rule about how writers are supposed to punctuate however in formal prose to reflect its status as a conjunctive adverb. At the same time, it’s generally recognized that students and other writers often struggle to master this rule, and “errors” with however persist even in more formal writing that has not been strictly edited (and occasionally that has). We could continue to view all of these examples of coordinating however as mistakes—as the product of writers who don’t know better and simply need to master the rule. Or, as our article does, we could examine the consistency in the “error.” Indeed, consistent “errors” in more colloquial contexts and/or with speakers or writers who haven’t yet fully internalized a prescriptive rule can signal language change at work. Has however become, at least for some speakers and writers of English, ambiguous in its syntactic categorization, a semi-coordinator in addition to a conjunctive adverb?

As we address this question, we advance the possibility that prescriptivism may be clouding our view of a linguistic change in the grammatical status of however. The prescriptive rule on however is pervasive, and it encourages us—non-linguists and linguists alike—to categorize variation in how however is punctuated as confusion about the rule and as a usage error. In the process, we can overlook the possibility that the source of the “confusion” might be a syntactic re-analysis of the word. Moreover, the case of however is exacerbated by the fact that the usage question is primarily visible in writing: we can’t hear with any accuracy whether a however between two clauses is functioning as a conjunctive adverb or a coordinator. But in the case of written language, the prescriptive rule that governs how we are to punctuate however by standard rules erases much of the potential evidence, at least in the large corpora in which language change can be most readily discerned.

As these observations suggest, the question of whether, at least for some speakers and writers, however has become a semi-coordinator in addition to a conjunctive adverb is a challenging one to answer. Most of the prose in available corpora is carefully edited according to house style guides, so coordinating uses of however are scarce at best: they have been edited out if the authors included any in the prose to begin with. We can, though, come at the question obliquely by considering correlate shifts: looking at other ways the use of however has shifted over time, in particular the word’s placement within the clause. The study we present here suggests that a change—or perhaps more accurately an expansion—in the syntactic status of however is a real possibility. Drawing on data from the Corpus of Contemporary American English (COCA) and the Corpus of Historical American English (COHA), this paper critically examines usage patterns to assess the extent to which however behaves like a conjunctive adverb and the extent to which its sentence placement could allow it be reinterpreted as a coordinator.

As part of this investigation, we are also tracking the interplay of usage and prescription. Recent scholarship suggests that prescriptivism can be one factor in language change (Anderwald 2012; Curzan 2014; Frazee & Wells 2015), and that to understand shifts in patterns of usage, especially in the written language but also in the spoken, we sometimes have to understand shifts in prescription. In this case, as we will discuss, the rise in prescription concerning clause-placement of however significantly correlates with a rise in the word’s frequency in clause-initial position. The subsequent rise in the prescription concerning the punctuation of however could thus be read as a response to significant confusion among speakers and writers about the status of however, due to its new clause-initial preference. To look at change in usage divorced from related prescription provides only part of the story; we gain a richer history of prescriptive rules—in this case, the rules focused on however—by looking at them alongside actual usage, both past and present. This perspective can reveal what grammarians are reacting to with the rules they are codifying. That said, an analysis of formal edited prose aimed at comparing actual usage and prescriptivism on a point of usage must come at such questions creatively, given the ways in which standardized, edited prose minimizes evidence of variation. We discuss this issue further in section 4.

With this study, we argue that the apparent “confusion” about whether however can serve as a clausal coordinator may be closely related to its increasing preference, over the past century-and-a-half, for clause-initial placement. The results raise the intriguing question—against many grammarians’ claims to the contrary—of whether however can function simultaneously as a conjunctive (or connective) adverb and as a coordinator. By looking at however in the context of the developments of a few other common conjunctive adverbs, we also evaluate the extent to which the behavior is idiosyncratic or part of a larger pattern.

First we talk about the terminology for the construction. Section 3 reviews the history of prescription on this construction, which sets up the corpus-based methods we needed to employ to come at this question. We then discuss the study’s findings and their implications for the grammatical classification of however and grammar/writing pedagogy.

2. Coordinators versus Conjunctive Adverbs

The conjunctive adverbial use of however is only about five hundred years old, according to the Oxford English Dictionary (OED). The word stems from the compound of the Old English hū (‘how’) and œfre (‘ever’). The fourteenth century turns up uses of however meaning ‘by whatever means’ or ‘to whatever extent,’ uses typically categorized as adjunctive (cf., Greenbaum 1969) and that remain in wide use to this day, as given in (6).

It is not until the early seventeenth century that the OED cites evidence of however used as the type of conjunctive adverb this paper addresses—a use the OED (s.v. however) describes this way: “Qualifying a sentence or a clause as a whole: For all that, nevertheless, notwithstanding; yet; = but at the beginning of a sentence.”

The OED entry hints at some of the confusion encountered by many present-day speakers, as it defines however with a mix of conjunctive adverbs (nevertheless, notwithstanding) and coordinators (yet, but). ² The OED’s definition ends with “= but at the beginning of a sentence,” suggesting that, in certain syntactic structures, however is equivalent to a traditional coordinator like but. Over the past two centuries, popular grammars and usage guides have sometimes categorized however as a conjunction (e.g., Harrison 1800; Greene 1850; Fowler 1965; Hacker 2004), although they distinguish conjunctions that can connect clauses from those that cannot, and posit that the ones that cannot function like adverbs. ³

Contemporary linguists typically draw a sharper categorical distinction between coordinators and conjunctive adverbs. ⁴ Some descriptive grammars admit to a blurry line between coordinators and conjunctive adverbs in some instances (e.g., so, yet), but however is not among those instances. All the descriptive grammars surveyed for this study agree that however is a conjunctive or connective adverb, not a coordinator or even a semi-coordinator.

By definition, a coordinator both syntactically and semantically “links” phrases and clauses of equal status, building coordinate syntactic structures (Greenbaum 1996:156; Biber et al. 1999:79; Huddleston & Pullum 2002:127). In comparison, a conjunctive adverb’s linking work operates only on the level of semantic connection. Conjunctive adverbs are “logical connectors that provide a link to a preceding sentence or clause” (Greenbaum 1996:146), making “explicit the relationship between two units of discourse” (Biber et al. 1999:765). Importantly, conjunctive adverbs, unlike coordinators, are optional within the clause structure (Biber et al. 1999:130). In other words, they do not perform necessary formal work but rather provide more semantic links. Removing a conjunctive adverb from a sentence affects neither what grammarians call its “truth value” nor the sentence’s syntactic integrity. Coordinators—at least in principle—cannot be similarly removed without affecting the integrity of the sentence (Greenbaum 1996:156; Biber et al. 1999:765).

Huddleston and Pullum (2002) echo much of this language about the work of conjunctive adverbs and then provide some useful specifications. Connective adverbs, which they categorize under “connective adjuncts,” “express the relation between the clause they introduce and what precedes it” (Huddleston & Pullum 2002:775), but they are not syntactically integral the way coordinators are. Huddleston and Pullum (2002) then suggest a set of properties—similar to Quirk et al.’s (1985:1289-1293) properties for a “central coordinator”—that characterize coordinators, including:

As this last feature captures, not all coordinators are prototypical—and not all conjunctive adverbs are prototypical either. As a result, the line between the two grammatical categories becomes less distinct.

Along these lines, and citing yet and so as examples, Huddleston and Pullum (2002:777) note that “the division between coordinators and connective adjuncts is not entirely clear cut”; they later explain: ⁵

so, in the sense ‘therefore, as a result,’ and yet, ‘nevertheless,’ appear in constructions where they are very different from coordinators, but they also have uses where the resemblances are such that they may be regarded as marginal members of the coordinator category. (Huddleston & Pullum 2002:1319)

What is a marginal member of a coordinator category? Geoffrey Pullum (2014, pers. comm.) explains that it describes words that have some properties of coordinators but not all, and the first two of the five properties above are the ones Huddleston and Pullum (2002:1320-1321) discuss specifically with so and yet. So and yet “normally occupy initial position,” which makes them more like coordinators than therefore or however (although we will show that “normal” behavior for these conjunctive adverbs seems to be shifting over time). While so and yet can occur with coordinators such as and, they more often occur without one (e.g., There was a bus strike on, so we had to go by taxi). Because they often appear alone between clauses in this way, Huddleston and Pullum (2002) argue that they are:

[. . .] just about indispensable elements of the construction. If we omit them the result is a mere juxtaposition of clauses rather than a coordination: it would therefore be misleading to treat the constructions here as asyndetic coordination [a construction where no coordinator is present]. It is more plausible to analyse so and yet here as markers of coordination. (1320)

In addition, so and yet occur with a range of coordinates, including finite verb phrases, seen in (7), and adjectives, as seen in (8).

Biber et al. (2002:227) look at this blurred boundary from another angle, showing where coordinators act more like conjunctive adverbs. They point out that the coordinators and, or, but, and nor, when they occur in turn-initial position in speech, act a lot like linking adverbials such as however. In other words, when a word like and occurs at the beginning of an utterance, it is expressing a semantic link to the previous utterance as opposed to a syntactic link. One can remove sentence-initial and without affecting the syntactic integrity of the sentence.

Now consider however in relation to these criteria for coordinators. The word however in a connective function often appears in sentence- or clause-initial position, although it doesn’t have to. This use of however cannot occur with a coordinator (*and however), so it meets that criterion. That said, this use of however cannot take an unlimited number of coordinates—but, as Huddleston and Pullum (2002) note, but, so, and yet do not meet this third criterion either. Coordinating however also connects only clauses.

Despite these mixed results in relation to the criteria for coordinators, however is regularly treated as a prototypical conjunctive adverb, as opposed to possibly (also) being considered a marginal coordinator. However regularly appears in lists of conjunctive adverbs, usually in the category of contrast or concession (e.g., Biber et al. 1999; Klammer et al. 2010). Huddleston and Pullum (2002) stand firm about its status as a conjunctive adverb—despite their sense that the line between connective adjuncts and coordinators can be hazy and their admission that items such as as well as and plus “in some uses have arguably been reanalyzed to become marginal members of the coordinator category” (2002:1293). In a footnote, the authors add:

However, whose meaning is similar to that of yet and but, also has uses where it behaves like a coordinator for speakers who accept such examples as %Other services have been expanded to fit the need, however the situation is still critical and %Please note that the costs are correct, however are subject to change prior to final payment [. . .] (Huddleston & Pullum 2002:1321)

The authors acknowledge such uses, but their indication that they are grammatical in some dialects only (marked by %) suggests that they do not accept these uses as having yet come close to the “marginal coordinator” status they grant to yet and so. In their chapter on punctuation, Huddleston and Pullum (2002:1742) admit to the routine nature of sentences in which however acts as a coordinator (e.g., “Your Cash Management Call Account does not incur any bank fees, however, government charges apply”)—but they assign this example sentence an asterisk to indicate ungrammaticality and assert that these sentences “would generally be regarded as unacceptable.”

Unacceptable indeed according to current style guides and grammar advice websites. But a more descriptive lens encourages us to admit that the above sentence might not appear ungrammatical or unacceptable to the many writers who feel inclined to use however this way. The lens of prescriptivism may be clouding our sense of what however is up to, dismissing as “errors” or “unacceptable” written uses that merit systematic analysis to see how speakers and writers understand however to function grammatically. And while however certainly can and often does act like a conjunctive adverb in terms of its flexibility in sentence placement, it also sometimes acts like a coordinator—and may even more so over time, as this study suggests.

Peterson (2009), in one of the few extended studies of however, argues that sentences like the one from the Harvard Review and Clean Eating, or the ones marked with % in Huddleston and Pullum’s (2002) grammar, reflect a change in the use of punctuation as opposed to a shift in the grammatical function of a conjunctive or connective adverb like however. He argues that however displays all the key syntactic properties of a connective adverb and should be seen as one; these run-on sentences with however, therefore, show a growing tendency to treat two closely related clauses as a single orthographic sentence, with an expanded use of the comma to do the connective work. We agree that punctuation alone cannot be used as a guide in syntactic analysis and as the basis of an argument for syntactic change. Although we use punctuation to delimit our data set, the evidence of change that this study provides focuses primarily on clause placement rather than punctuation. Further, and unlike Peterson (2009), we suggest that it is the function of the word, and not the function of the punctuation, that may be changing—that however itself may increasingly be behaving like a coordinator.

Peters (2004) briefly mentions this possible coordinator use of however in The Cambridge guide to English usage. She argues that coordinator however is largely a spoken phenomenon: “However has traditionally been associated with academic and discursive writing, but it is also heard now from time to time in conversation. Its role there is beginning to affect its grammar” (Peters 2004:255, emphasis added). ⁷ Our study explores Peters’ (2004) observation that shifts in however use may be influenced by the spoken language and suggests the factors involved in any such shifts with however are not restricted to the spoken. We also recognize that written evidence for “how however works” is significantly affected by widespread prescription on its usage, as we will discuss in section 3, before turning to methods.

3. A History of Prescription on However

This section provides an overview of the progression of twentieth-century prescriptive energy focused on “correct” and “unacceptable” uses of however. One focal point of this energy is a now often forgotten rule about the correct placement of however within the clause; a second focal point is the well-known rule about punctuating however with a preceding semicolon. For this study, we centered our inquiry on well-known, popular grammar books that went through multiple editions, as well as a selection of competing guides. The summary is not designed to be a systematic survey of prescriptivism on however. Even so, a broad understanding of the history of these rules provides key context for understanding changes in patterns of actual usage in written data over the past two centuries. Further, the data from this study suggest a complex interplay of prescription and usage.

The first rule may now seem irrelevant, but it becomes relevant for our methods in showing how prescription and usage have interacted over the decades. Though now largely moribund, this rule discourages the use of clause-initial however, in its conjunctive adverb function, altogether. It is potently stated in a number of the most well-known American and British guides up through the 1990s. In one of its earliest manifestations, William Strunk in The elements of style (1918:36) advises that however, “in the meaning nevertheless,” is “not to come first in the sentence or clause.” In the now classic 1959 version, revised and expanded by E. B. White, this prescription is worded even more powerfully: “[a]void starting a sentence with however when the meaning is ‘nevertheless.’ The word usually serves better when not in first position” (1959:39). Gower’s revision of Fowler’s A dictionary of modern English usage (1965:188) and William Zinsser (1976:98) make the same recommendation. Partridge and Greet’s Usage and abusage (1942:154) advises that however “comes [. . .] after the first significant unit.” Garner (2016:472) calls clause-initial use a “stylistic lapse.” In a more descriptive key, the 1977 Warriner’s English grammar and composition: Complete course suggests that the term’s mid-clause use is most logical, and the guide describes however as “rarely used at the beginning of a clause” (1977:538).

As we will show below, our study complicates Warriner’s (1977) claim that clause-initial however was rare in the 1970s. The data show instead that the popular growth of this rule—specifically between the twenties and seventies—actually corresponds to a dramatic rise in clause-initial placement of however in published material. Many usage guides—Strunk and White (1959) among them—explain this proscription as a guard against confusion between the term’s subordinator function and its sentence-modifying function (e.g., between adverbial uses like “However the steak was cooked, I ate the whole thing” and conjunctive adverbial uses like “However the steak was cooked, and I ate the whole thing”). ⁸ We speculate, though, that this proscription against clause-initial however may have functioned at least in part as a guard against sentence-initial, more coordinator-like uses of however. By discouraging conjunctive adverb however from moving into sentence-initial position, it would not be so easily confused with a coordinator.

Yet, as is probably intuitive to most contemporary writers of English, the strength of this rule has faded considerably in recent decades. Peters’ The Cambridge guide to English usage (2004:255) succinctly summarizes: “there is no basis for suggesting that contrastive however should not appear at the beginning of a sentence.” American heritage dictionary (online) notes the prescription, but also shows it to be losing considerable ground. In the 2015 survey, only 10 percent of the dictionary’s usage panels reported always following this rule, 30 percent usually or sometimes, and 60 percent rarely or never. Further, the 2015 survey shows that the corrective impulse around this rule has also faded: fewer than 10 percent said that they judge sentence-initial however as an unacceptable use of the term. In other words, only one in ten would edit this use out of someone else’s prose; all others find clause-initial however either “somewhat” or “completely acceptable” (2015: s.v. however).

This increasing tolerance for clause-initial however arguably puts even more pressure on the second and more popular rule concerning however, this about how to correctly punctuate the word with a preceding semicolon. The use of the semicolon makes it clear that however is not a coordinator, even when the term occurs between two clauses. In the 2005 edition of Style, Joseph Williams (2005:230-231) notes that “even well-educated writers often incorrectly use a comma to end one grammatical sentence when they begin the next sentence with however.” It is worth noting that this admission about “well-educated writers” distinguishes Williams from Bryan Garner (2016:472), who, as quoted above, attributes such tendencies only to “informal writing, student papers, and the like.” The recognition that the well-educated “slip up” seems to provide all the more reason for Williams to restate the rule in his guide. He even characterizes such “errors” as a “special problem with semicolons and however” (Williams 2005:230)—implying that, compared to its fellow conjunctive adverbs, this term’s grammatical status becomes particularly slippery when it takes that clause-initial position where only proper punctuation can provide a clarifying fix. As it still is for many editors and instructors of English, the insertion of the semicolon became the refrain.

The descriptive Encyclopedia of English (Zeiger 1978:152) notes an overwhelming propensity among English teachers to “before however [. . .] insist upon the semicolon” in sentences containing “coordinate clauses.” Warriner and Griffith (1977:538) advise writers to “use a semicolon between independent clauses joined by words such as for example, for instance [. . .] however [. . .].” Hacker (2004:288) recommends that writers “use a semicolon between independent clauses linked with a transitional expression,” including however. And Garner (2016:472), as referenced above, notes that “however often plays a role in run-on sentences. [. . .] One cure, of course, is a semicolon[. . .] .” ⁹

The general outlines of this rule can be found in usage guides dating back to the twenties, where, although often framed as a more general prohibition of the comma splice, however tends to function as this error’s prime example (e.g., Frazee & Wells 1921; Conlin 1961): in one instance, Sterling Leonard’s 1932 Current English usage unequivocally states that “there is little room for doubt; however should take the semicolon” (1932:20).

In no usage guides did we find a tolerance for coordinator however—for a clause-initial position between two complete clauses and punctuated only with one or two commas. Among usage panels, the prescription has remained strong. Drawing on a 1932 National Council of Teachers of English survey about the rules of “standard punctuation,” the Encyclopedia of English reports that connective however would have been deemed acceptable only by a small minority of “dissenting voices” (Zeiger 1978:152). The 2016 American heritage dictionary survey of its usage panel provided the sentence “Main Street will be closed to traffic for the parade, however, the stores along it will remain open” and found that a full 86 percent would judge such a use of coordinating however unacceptable. (That 14 percent of the Usage Panel rated the sentence in the realm of acceptable is notable, though; s.v. however.)

Two conclusions can be drawn from this history. One is that, with all of this prescription in circulation, it is completely reasonable that most editors would correct any use of “unacceptably” punctuated connective however in prose to be published, often with the simple exchange of a semicolon or period for a comma. This editing raises significant hurdles for studying coordinating however in corpora of published writing and transcribed speech, as we address in section 4. The second is the complex interplay that exists between these shifting prescriptions and actual shifts in use, a dynamic in which the pervasive power of prescription can make the difference between real change and “error” difficult to discern.

4. Methods

As previously noted, the immediate challenge of conducting a corpus-based investigation of coordinating however is that there are very few examples in the published writing collected in most standard corpora. If such instances existed at all in an earlier draft, they were probably edited out. Further, transcripts of spoken language tend to adhere to prescriptive conventions for how to punctuate however; after all, how can you hear if a clause-initial however is preceded by a period, semicolon, or comma? That doesn’t mean, however, that we cannot study factors that might encourage the reanalysis of however as a coordinator. For example, how often is however used clause non-initially (as we did in the previous sentence) versus clause-initially, where it could be interpreted as a coordinator? The distinction between “clause initial” and “clause non-initial” serves our purposes because only in clause-initial position can however be interpreted as a coordinator. In non-initial position, whether that be after, for example, a prepositional phrase, as seen in (9), or between the subject and the main verb, as seen in (10), or at the end of a clause, however must be interpreted as a conjunctive adverb (see Lenker 2011 for a more detailed discussion of different medial positions for adverbial connectors).

To come at the question of coordinating however obliquely, we analyzed uses of however as a conjunctive adverb in edited written American English over the past two hundred years. In order to analyze both historical and current uses, we relied primarily on the Corpus of Historical American English (COHA) and the Corpus of Contemporary American English (COCA). ¹⁰ COHA, with over 400 million running words of print text from 1810–2009, provides data on overall trends in the word’s frequency over time. COCA, with over 450 million running words from 1990–2017 covers five registers: spoken, fiction, popular magazines, newspapers, and academic texts. As a result, COCA allowed us to track very recent change and perhaps more importantly, the distribution of the word’s use across different registers.

First we needed to establish a search method that would track, on a large scale, any changes in the use of conjunctive adverb however both over time and over registers, as they relate to both overall frequency of use and sentence-placement trends. Of course, corpus-based searches for however as a conjunctive adverb are complicated by the fact that however can also function as a subordinating conjunction or subordinator, as seen in (11), and as a non-conjunctive adverb, a seen in (12).

However, random sampling of the edited prose in COCA confirmed, across its five registers, two highly consistent punctuation trends that helped us limit our data set to conjunctive adverb uses of the term, with only minor noise in the data. First, conjunctive adverb uses of however are almost always followed by a comma, be that at the beginning or in the middle of a clause or sentence (error rate of only 1-3 percent in a random sampling of the data); at the same time, subordinator or non-conjunctive adverb uses of however are never followed by a comma (at an error rate of 0 percent). ¹¹ Not surprisingly, such punctuation use is in accordance with the recommendations of many contemporary usage guides (Cioffi 2015:189; Garner 2016). Second, this random sampling confirmed that coordinator however is almost never found in the carefully edited prose that constitutes COCA and COHA. We, at least, found no instances of it in our random sampling; we found only a few in the spoken register. (Our recent finding from The New York Times, example (5), thus stands as a very rare exception.)

For the overall frequency searches for however as a conjunctive adverb, we therefore searched for “however,” with a non-case-sensitive search, which retrieves both sentence- or clause-initial and sentence-/clause-non-initial uses of the word. We recognize that the presence of the comma slightly undercounts the number of conjunctive adverb uses, as it misses both those few instances that do not include a comma, as seen in (13), and instances of clause-final however, as seen in (14). (Note that we do include sentence-/clause-final however in our detailed analysis of sentence placement patterns.)

But the presence of the comma also kept us from over-reporting by eliminating from our results all other syntactic uses of however. Our random sampling gave us further confidence that our searches were not retrieving instances of “coordinator however” (which in colloquial use is often surrounded by commas) since we found that such instances rarely, if ever, turned up in COCA’s edited prose. Given the size of the numbers in COHA and COCA (COCA, for example, reports over 170,000 instances between 1990 and 2017), the findings from our random samplings (an average error rate of 0-2%), and the consistency of the search across time and registers, we believe that these search methods provide an accurate sense of overall patterns—and a sense which could not be achieved through any other means.

Punctuation played a critical role in searching for sentence-placement patterns as well. In order to capture uses of sentence-/and clause-initial however (which for the rest of the article we refer to together as “clause-initial however”), we searched in COHA and COCA for [. However,] and [; however,]; and in order to capture clause non-initial uses of the conjunctive adverb, we searched for [, however,] and [, however .]. Again, we recognize that this search method did omit some uses of the conjunctive adverb, but punctuation patterns proved a highly reliable way to find the conjunctive adverb, given the current style guidelines with which we can assume the corpus’ prose was edited. As the American heritage guide to contemporary usage and style (2005:230) summarizes: “[w]hen however is used as a conjunctive adverb, its clause must either follow a semicolon or be a new sentence. In these uses, however is usually set off by one or more commas.” While we have found scattered instances of coordinator however surrounded by commas in non-corpus-based sources like the Harvard Review, these remain rare enough in strictly edited prose to remain a non-factor in the large-scale corpus searches of this study. We also did not try to categorize the discourse functional uses of medial however (cf., Lenker 2011); our focus here is on the frequency of clause placement with the basic contrast of initial versus non-initial.

5. Historical and Current Use of However

As we will show below, however is one of the more popular conjunctive adverbs in written prose, especially academic prose (see also Biber et al. 1999:887). That said, it is not nearly as popular as it used to be. Based on the search methods described above, which limits the data to instances of the conjunctive adverbial use of however, we find overall that the written use of however has dropped dramatically over time, according to COHA data (see Table 1 and Figure 1). From a highpoint of 452.28 occurrences per million words (pmw) in the 1820s, the frequency has been cut in half by the first decade of the twenty-first century, appearing 217.81 times per million words from 2000–2009. These changes in frequency have high statistical significance (χ² = 86.71, df 4, p < .0001), with the 1820 decade being 35 percent higher than average and the 2000 decade being 35 percent lower.

OPEN IN VIEWER

Table 1.

Frequency of Conjunctive Adverb However 1820–2009 (COHA)

Period	Raw Frequency	pmw
1820–1829	3133	452.28
1850–1859	6035	366.39
1900–1909	7056	319.31
1950–1959	7797	317.66
2000–2009	6440	217.81

OPEN IN VIEWER

Figure 1.

Normalized Frequency of However 1820–2009 (COHA, pmw)

This decline in written use could reflect the larger trend of colloquialization in the language—or, as Leech et al. (2009:244) describe it, a “negative manifestation of colloquialization”: in other words, the reduction of a feature associated with formal writing as part of the increase of more spoken features into writing. Previous studies of colloquialization have demonstrated, for instance, an increased use in writing of contractions, as well as the rise of informal semi-auxiliaries such as have to/hafta and going to/gonna (see Leech et al. 2009 for a useful summary). While however may not have been as formal in the nineteenth century, increasing genre differentiation over the course of the nineteenth century led to the much higher frequency of conjunctive adverbs generally in scientific and academic writing that is characteristic of the twentieth century (Grund & Smitterberg 2014). The conjunctive adverb however now heavily favors written genres, especially formal ones; for example, in COCA it appears ten times more often in the academic register than it does in the spoken register (844:81 words per million; see also Biber et al. 1999:880-882). So the fact that writers are using however less over time does seem to correlate with the move toward more colloquial language.

Nonetheless, however remains a relatively popular conjunctive adverb. In 2000–2009 data in COHA, however followed by a comma occurs at a rate of 217.81 pmw. To provide a comparison, as can be seen in Table 2, however occurs almost three times more often than the conjunctive adverb thus (76.81 pwm), more than four times more often than therefore (48.16), and almost nine times more than nevertheless (23.91). Interestingly, all three of these more formal conjunctive adverbs have also been declining in frequency since the beginning of the twentieth century, and as early as the mid-nineteenth century for therefore (see also Biber & Gray 2010). The conjunctive adverbs therefore and thus have diminished in frequency since the beginning of the nineteenth century seven- or eightfold (see Table 2). In contrast, a less formal, more colloquial conjunctive adverb like plus has been rising in frequency. ¹² In sum, while however’s overall use is declining, it remains popular compared with these three relatively frequent conjunctive adverbs, and the patterns of decline and growth for all four conjunctive adverbs in COHA correlate with the broader trend of colloquialization in written language.

OPEN IN VIEWER

Table 2.

Common Conjunctive Adverbs 1820–2009 (COHA, pmw)

Conjunct	1820–1829	1850–1859	1900–1909	1950–1959	2000–2009
Nevertheless	55.29	54.82	56.61	53.17	23.91
Therefore	417.49	276.48	180.83	127.40	48.16
Thus	562.13	611.17	326.60	187.58	76.81
Plus	0	0.12	0.09	0.53	13.60

More significant for this study than the story of decline are the results for shifts in the clause placement of however over time. Uses of clause-initial however have dramatically risen, such that in the COHA data this use is three times more common at the start of the twenty-first century than it was in the mid-nineteenth century, as seen in Table 3 and Figure 2. This trend is statistically significant (χ² = 44.01, df 4, p < .0001). Over the same period, the frequency of clause non-initial however has been more than cut in half, again a trend that is statistically significant (χ² = 87.85, df 4, p < .0001). This reveals, over the century and a half—and against the proscription against clause-initial use noted in our history above—an increasing, if somewhat uneven, tendency for however to be used at the beginning of sentences and clauses. ¹³

OPEN IN VIEWER

Table 3.

Sentence-placement Preference of Conjunctive Adverb However 1850–2009 (COHA, pmw)

Period	Clause non-initial however ([, however,] + [, however .])	Clause-initial however ([. However,] + [; however,])
1850–1859	335.75 (335.55 + 0.2)	21.61 (17.79 + 3.82)
1900–1909	279.44 (278.94 + 0.5)	32.13 (30.77 + 1.36)
1950–1959	223.31 (222.78 + 0.53)	79.60 (77.69 + 1.91)
2000–2009	137.54 (136.84 + 0.7)	62.57 (59.15 + 3.42)

OPEN IN VIEWER

Figure 2.

Normalized Frequency of Clause-initial However 1850–2009 (COHA, pmw)

These trends can be put in conversation with concurrent prescriptions in circulation. As pointed out above, a number of grammarians began to strongly enforce the rule prohibiting clause-initial however long after written usage had begun to turn a significant corner in terms of the rising frequency of clause-initial however. Warriner and Griffith’s (1977) recommendation about clause-initial however, which they even make on descriptive grounds, provides a case in point. COHA shows that in the decades between 1950 and 1980, during which Warriner and Griffith (1977) wrote and revised their guides, clause-initial however came to be used about 50 percent as frequently as clause-medial however, with the popularity of this use having steadily increased since earlier in the century (see Table 4). Our data thus challenge their 1977 claim that conjunctive adverb however is “rarely used at the beginning of a clause” (Warriner & Griffith 1977:538)—unless half-as-often qualifies as “rare.”

OPEN IN VIEWER

Table 4.

Sentence-placement Preference of Conjunctive Adverb However 1930s–1970s (COHA, pmw)

Period	Clause non-initial however ([, however,] + [, however .])	Clause-initial however ([. However,] + [; however,])
1930–1939	256.03 (255.38 + 0.65)	60.73 (59.51 + 1.22)
1940–1949	227.35 (226.75 + 0.6)	64.36 (63.21 + 1.15)
1950–1959	223.31 (222.78 + 0.53)	79.60 (77.69 + 1.91)
1960–1969	204.93 (204.24 + 0.69)	87.75 (84.91 + 2.84)
1970–1979	196.58 (195.88 + 0.7)	83.89 (80.87 + 3.02)

Strikingly, this strongly worded prescription about sentence-initial placement began to take popular hold just as this clause-initial use was reaching its apex in the fifties and sixties. The strength with which this rule was stated over a number of decades—as with Partridge and Greet’s (1942) early editions of Usage and abusage, or White’s revisions of Strunk’s Elements of style (1959)—appears, therefore, to be a reaction against real growth in clause-initial placement. This rule’s eventual decline (captured in the 2015 usage survey of the American heritage dictionary Usage Panel) also signals an acceptance of the longevity of the usage preference for clause-initial placement. It is also possible that the dying of this sentence-placement rule has reinforced the power of the second rule concerning the mandatory use of the semicolon. Setting these prescriptions alongside usage trends suggests that over the last sixty years, grammarians have seen a need to control however’s coordinator-like tendencies.

To be clear, the COHA data still indicate that clause non-initial however remains twice as frequent as clause-initial however, and this, we can surmise, is because COHA data draw exclusively on edited prose; but the COCA data tell a different story about these shifting ratios in placement preference, as we discuss below. The COCA data indicate a stronger conflict between prescription around this term and real use.

Important to remember is that, unlike COHA, COCA includes spoken data (in a formal register, but spoken nonetheless) along with both formal and less formal registers of written language. According to COCA data, the trend over the past twenty-two years shows clause-initial however steadily overtaking clause non-initial however, as clause-initial use rises in frequency and clause non-initial use slowly falls. Strikingly, the most recent period of 2010–2014 and 2015–2017 shows clause-initial use of however becoming slightly more frequent than the clause non-initial use (see Table 5 and Figure 3). When broken out by year, these changes also prove statistically significant (see Appendix A). In other words, the tipping point favoring clause-initial however has already occurred.

OPEN IN VIEWER

Table 5.

Sentence-placement Preference of Conjunctive Adverb However 1990–2017 (COCA, pmw)

Period	Clause non-initial however ([, however,] + [, however .])	Clause-initial however ([. However,] + [; however,])
1990–1994	223.94 (210.93 + 13.01)	110.01 (102.96 + 7.05)
1995–1999	193.76 (182.32 + 11.44)	112.77 (104.68 + 8.09)
2000–2004	180.40 (169.89 + 10.51)	119.87 (111.04 + 8.83)
2005–2009	149.31 (140.52 + 8.79)	115.85 (105.58 + 10.27)
2010–2014	124.67 (116.9 + 7.77)	127.64 (112.63 + 15.01)
2015–2017	122.42 (115.34 + 7.08)	131.48 (117.24 + 14.24)

OPEN IN VIEWER

Figure 3.

Comparative Normalized Frequency of Clause Non-initial and Clause-initial However 1990–2017 (COCA, pmw)

These data suggest that about half the time current speakers and writers of American English encounter however, the term is in clause-initial position—and this, significantly, is the very syntactic placement where the term can readily be interpreted as a coordinator. The data also indicate that over just the past twenty years, usage has changed dramatically in terms of the sheer frequency with which speakers and writers encounter clause-initial however: it would have been about one in three instances in the early 1990s, rather than one in two in more recent years.

In the COCA data, spoken language comes to favor clause-initial however by 2012, but not as strongly as fiction, all of which could be read as evidence for the colloquial nature of clause-initial placement. The diachronic trend toward clause-initial however is strongest in spoken language, fiction, and academic prose; newspapers and magazines in COCA still heavily favor clause non-initial however. As Tables 6 and 7 show, in the spoken register of COCA, clause-initial uses of however overtake clause non-initial uses in 2000–2004, and by 2015–2017 these uses are favored by a ratio of more than 2:1; in academic prose the balance tips toward clause-initial uses of however in 2005–2009, and by 2010–2014 and 2015–2017, these uses are favored almost 2:1. The increasing academic preference for clause-initial uses complicates an argument for clause-initial however as a straightforwardly colloquial feature.

OPEN IN VIEWER

Table 6.

Sentence-placement Preference of Conjunctive Adverb However in the Spoken Register 1990–2017 (COCA, pmw)

Period	Clause non-initial however ([, however,] + [, however .])	Clause-initial however ([. However,] + [; however,])
1990–1994	12.98 (12.14 + 0.72)	7.14 (6.81 + 0.33)
1995–1999	8.08 (7.49 + 0.59)	7.20 (6.68 + 0.52)
2000–2004	8.68 (8.03 + 0.65)	9.90 (9.63 + 0.27)
2005–2009	5.33 (5.27 + 0.6)	6.57 (6.31 + 0.26)
2010–2014	3.97 (3.59 + 0.38)	6.66 (6.58 + 0.08)
2015–2017	2.80 (2.62 + 0.18)	6.04 (5.99 + 0.05)

OPEN IN VIEWER

Table 7.

Sentence-placement Preference of Conjunctive Adverb However in the Academic Register 1990–2017 (COCA, pmw)

Period	Clause non-initial however ([, however,] + [, however .])	Clause-initial however ([. However,] + [; however,])
1990–1994	96.91 (93.9 + 3.01)	68.98 (63.9 + 5.08)
1995–1999	93.64 (89.80 + 3.84)	70.48 (64.55 + 5.93)
2000–2004	83.09 (79.87 + 3.22)	77.69 (70.48 + 7.21)
2005–2009	66.20 (63.63 + 2.57)	78.22 (69.63 + 8.59)
2010–2014	44.54 (43.25 + 0.1.29)	91.39 (77.76 + 13.63)
2015–2017	50.20 (48.98 + 1.22)	88.54 (75.69 + 12.76)

Overall, the shift in sentence placement for however, while it may provide evidence for ongoing colloquialization, is even more important because it arguably sets up the conditions for syntactic reanalysis. Syntactic reanalysis requires a grammatically ambiguous construction, and it involves the reinterpretation of the underlying syntactic structure without necessitating the modification of the surface structure (Harris & Campbell 1995). In order for syntactic reanalysis to occur, only some but not all surface manifestations need to be open to the reinterpretation, as is the case here. In clause non-initial position, as we have already discussed, however can only be interpreted as a conjunctive adverb. In clause-initial position, in contrast, a conjunctive adverb use of however could readily be reanalyzed as a coordinator. It is semantically similar to but, which is a coordinator; and its growing preference for coordinator position may encourage speakers to view it as such. It may be in the lead of this kind of reanalysis because of its relative frequency compared with the semantically similar nevertheless.

As a result, we are arguing that however at this point has similarities with so and yet, which Huddleston and Pullum (2002:1319) categorize as “marginal coordinators.” Without question, however still often functions as a conjunctive adverb, and necessarily so when it appears in clause-medial position. But the data from COHA and COCA suggest that it may also be time to accept the possibility that it can function for many speakers as a coordinator, too. Peters (2004:256), who also acknowledges that however is sometimes used as a conjunction in spoken English, notes that prescriptions against this “shift to conjunction” are based on “reasons unclear, except that [such a shift] would enlarge what is usually taught as a closed set of conjunctions.”

6. Conclusion

The data presented here ask us to take seriously the possibility that however now functions as a conjunctive adverb and a “marker of coordination,” to quote Huddleston and Pullum (2002:1320) about yet—a marginal or semi-coordinator. In other words, speakers and writers are not simply “confused” about how to use the conjunctive adverb however; these speakers and writers have reinterpreted the term’s grammatical status in some constructions such that it functions as a coordinator between clauses. This argument implies that a clause-initial however could be functioning syntactically in two different ways. We recognize the counterargument that as long as however can be moved to clause non-initial position, it cannot be seen to function as a coordinator; it is simply a conjunctive adverb appearing clause initially. In current usage, in the constructions where however appears clause initially and could be interpreted as linking clauses syntactically (e.g., She is rarely on time, however, yesterday she was early), it could at least theoretically be moved to clause non-initial position, where it clearly functions adverbially (e.g., She is rarely on time; yesterday, however, she was early). But the fact that clause-initial however “could” be moved to clause non-initial position does not address the increasing preference of speakers and writers to place the word in clause-initial position. The hypothesis that however could now be categorized as a marginal coordinator helps explain this trend in sentence placement. In other words, we may have mutually reinforcing processes at work: the increasing use of clause-initial however may encourage speakers to interpret however as a coordinator; and the interpretation of however as a coordinator requires that the word be used clause initially.

A question that deserves further inquiry, therefore, is whether we can distinguish conjunctive adverb however from semi-coordinator however in clause-initial position. It could be interesting to pursue prosodic analysis of spoken language to see if differences emerge that might signal different syntactic functions. That said, the coordinator but allows a range of prosodic patterns in clause-initial position. In the written language, standardization of punctuation makes it much more difficult to tell: even writers for whom the term however may function as a coordinator have been taught not to punctuate it that way. Thus, the orthographic rule may mask the way however is functioning; the word’s function does not determine the orthographic rule, at least in most formal edited prose.

What happens, then, if we let the prescription against coordinator however go—if we follow the language trends? Usage guides would certainly have an easier time admitting to multiple classifications, instead of existing in the tangle in which many now find themselves: presenting however as a term synonymous with but, which often appears clause-initially, yet that needs to follow different orthographic conventions when it appears clause initially in a compound sentence. As we detail in section 2, the semicolon “cure” would not be the first prescriptive rule about however to fall by the wayside. The proscription of clause-initial however now seems old-fashioned to many writers, even if they are vaguely familiar with it. And as clause-initial however rises in relative frequency to clause non-initial however, it makes sense to track the health of the rule that requires it to remain a conjunctive adverb in this position. If the force of the prescription eases—and it is now at a very strong ebb—usage trends indicate that evidence of the coordinator function may become even more prevalent. These are interacting forces at play, between directives about usage and uptakes by users. In order to fully understand the trajectories of usage for a term like however—which is currently highly regulated—we need to consider the power of prescription, especially on written usage (Curzan 2014).

For any of us who act as grammatical gatekeepers in our roles as copy editors and/or writing instructors, we may not yet be ready to allow however through the gate as a coordinator with just a comma and no semicolon; the risk that the writer (or the copy editor) will be judged negatively remains too high, given the continuing strength of this prescription, not to mention our own commitment to however’s status as non-coordinating. We propose that recognizing the ways that the term however appears to be in flux, though, could lead to richer pedagogical and grammatical conversations about how we rhetorically use however: why and in what circumstances, on the level of both register and sentence placement, however does the work that we see it doing. And this is a conversation, we suggest, that should resist the easy answer that the coordinator use of contrastive however is simply unacceptable or wrong, as so many grammarians and English instructors now insist.

In fact, the specific case of however reveals the extent to which researchers’ own grammatical classifications, which we tend to think of as descriptive, can in and of themselves be prescriptive. As long as we assume however can function only as a conjunctive adverb, then all the writers who are “mis-punctuating” it must be seen as novices, not yet fully in control of standard punctuation conventions, rather than as innovators who have reinterpreted however—on very logical grounds—such that it could be considered a marginal coordinator. This case study also shows how prescriptive rules about usage both respond to language change and can taint evidence of that language change in written documents. It is our hope that both research into usage trends, and instruction in usage, can account more fully for the complex forces at work in language change, especially by tracking not only change but the ideologies informing the rules by which we conduct our research and draw our conclusions. Nowhere is this more important than with those instances of usage that—like however— are particularly freighted with formal prestige and receive sustained corrective energy.