3. Data and methodology

The approach adopted here is largely based on that taken in historical-sociolinguistic studies investigating the effects of prescriptivism on language change, referred to as precept vs. practice (Auer & González–Díaz, Reference Auer and González–Díaz2005). The method consists of comparing the treatment of usage features in normative grammars, i.e. precept, to patterns of use observed in actual language use, i.e. practice (e.g. Dekeyser, Reference Dekeyser1975; Auer & González–Díaz, Reference Auer and González–Díaz2005). Similarly, in studies of the influence of usage guides on actual usage (e.g. Albakry, Reference Albakry2007; Peters, Reference Peters2014; Ebner, Reference Ebner2017; Kostadinova, Reference Kostadinovaforthcoming), usage guides pronouncements are contrasted with patterns of actual use of selected language variants in corpora. In the same vein, this paper uses precept data from American usage guides, and corpus data from the Corpus of Historical American English (COHA; Davies, Reference Davies2010–) and the Corpus of Contemporary American English (COCA; Davies, Reference Davies2008–) in order to compare the treatment of literally as a usage problem with its actual use. For evidence of speakers’ attitudes towards the variation in the use of literally, I will draw on data from interviews with native speakers of American English I conducted in 2014. In what follows, I will discuss each of these datasets in more detail.

3.2. Corpus data

The data on the patterns of actual use consists of occurrences of literally extracted from the full-text data of COHA and COCA. Because of the necessity for a manual disambiguation of the occurrences of literally, a representative sample was used for the analysis, obtained by selecting every fifth occurrence of literally from the total number of occurrences; details are given in Table 1.

Table 1: Corpus data for literally retrieved from COHA and COCA

The occurrences in the samples were then disambiguated on the basis of a tripartite classification of uses, adapted from the five uses identified by Powell, discussed in Section 2 above; the three categories I distinguish are basic use, dual use and nonliteral use. Basic use refers to the primary uses of literally, exemplified in (1) above. Dual use comprises three of the uses of literally identified by Powell (Reference Powell, Lehrer and Kitay1992: 341–4), which I discussed above, viz. literally with expressions denoting ‘extreme cases’ (example [2]), literally in ‘dual readings’ (example [3]) and literally as ‘semantic innovator’ (example [4]). These are all cases in which literally is used to force a literal reading of an expression which is otherwise likely to be interpreted nonliterally. Leaving literally out in all these cases would affect the interpretation of the expression it modifies. In the context of my analysis, I refer to all three of these as dual use, because in all of them both the conventional and the literal meanings are operative. The third category, nonliteral use, is the use of literally with nonliteral expressions, exemplified in (5). These uses violate the basic, propositional meaning of literally, thus maximising its nonliteral, pragmatic force. The decision to conflate three of the uses identified by Powell under the category dual was made primarily because I am interested here specifically in the occurrence of the primary and the nonliteral uses of literally, which is important for the question of the relationship between usage guides treatment and actual use (but see Kostadinova, Reference Kostadinovain preparation).

4. Literally in the usage guides

From the point of view of its discussion in the usage guides, literally is a typical 20^th-century usage problem, as shown in Figure 1.

Figure 1. Number of usage guides which cover literally from the total number of usage guides analysed per decade

The classification of these entries into the three mutually exclusive categories acceptable, restricted and unacceptable established in Section 3.1 reveals that this feature is largely considered unacceptable. Of the 32 entries, 24 were classified as unacceptable, meaning that the authors accept only literal uses of literally. The remaining 8 entries were classified as restricted, i.e. the authors either accept the different variants of literally, including the nonliteral use, in certain less formal contexts, or they are vague in their treatment, and do not express any attitudes to the acceptability of the feature overtly. In terms of changes in treatment, then, the pattern in Figure 2 shows that there has been no change in how literally is perceived over the years from the point of view of correct usage. The presence of a small number of restricted entries is more likely to be an effect of variation within the genre of usage guides than an indicator of a change in precept, as less prescriptive usage guides tend to acknowledge patterns of language variation and change without overt negative judgements.

Figure 2. Treatment of literally in the usage guide entries per decade

The number of positive and negative attitudes in the entries shows a striking contrast, with negative attitudes being far more common than positive ones. This is hardly surprising, given the overall treatment of literally. Positive attitudes are rare, and occur only in two guides, which, for instance, acknowledge that literally is ‘neither a misuse nor a mistake’ (Gilman, Reference Gilman1989: 607), but a logical development within a process of language change. Furthermore, instead of deriding those who use literally in not strictly literal uses, one entry (Peters, Reference Peters2004: 326) recognises the subjective and expressive nature of this use. Despite this more balanced account, these two guides still advise language users to exercise care in using literally for emphasis. Negative attitudes are far more prevalent. From this perspective, literally is considered a ‘misuse’ (Copperud, Reference Copperud1980: 231; Johnson, Reference Johnson1991: 365), which is ‘intolerable’ (Bierce, Reference Bierce1909: 42) or ‘superfluous’ (Pinker, Reference Pinker2014: 276). Its use is associated with confusion about its meaning (O'Conner, Reference O'Conner1996: 85), or absurdity of expression (Ebbitt & Ebbitt, [1939] Reference Ebbitt and Ebbitt1978: 548). The claim that this use of literally prevents the reader from understanding what is meant (Clark, Reference Clark2010: 148) and may evoke ‘ludicrous’ images as a result (Stilman, Reference Stilman1997: 39; Pinker, Reference Pinker2014: 276) is also found in the entries. As an intensifier, literally is seen as ‘excess baggage’ (Copperud, Reference Copperud1980: 231) that ‘should be avoided’ (Evans & Evans, Reference Evans and Evans1957: 280).

In addition to the predominance of negative attitudes in the usage guides analysed, two other aspects of the treatment of literally are worth addressing. The first is that the entries sometimes contain observations about the use of literally which are in line with descriptive accounts, even if the advice offered in the entries may be prescriptive. Observations that literally is ‘acceptable only when it serves to show that an expression usually used metaphorically is to be taken at its word’ (Bryson, Reference Bryson1984: 93) or that it ‘is used more legitimately to emphasize the reality of facts or events that could be mistaken as figurative’ (Sutcliffe, Reference Sutcliffe1994: 79) illustrate that in some cases usage guides may contain descriptively relevant information about language use (in the context of historical-sociolinguistic studies this argument has been made before; see e.g. Poplack & Dion, Reference Poplack and Dion2009; Tieken–Boon van Ostade, Reference Tieken–Boon van Ostade2011; Anderwald, Reference Anderwald2012). The second aspect is that there seems to be a lack of understanding of the function of literally, especially in dual use contexts. Usage guide writers insist that literally is only acceptable in its primary, literal use and criticise intensifying uses of literally. However, in actual usage, the intensifying function of literally is combined with its literal meaning.

5. Literally in the corpus data

Turning to the results from the disambiguation of the occurrences of literally in the sample, Figure 3 shows that the most common use is the dual use, followed by the basic use, with the nonliteral use being substantially the least frequent. This situation, moreover, does not appear to have changed drastically over the last two centuries. Although the proportion of dual use seems to be somewhat larger in COCA than in COHA, at the expense of the basic use, the difference is still small. It is difficult to know whether this difference is indicative of the rate of change in the distribution of the different uses of literally, or whether it is merely the result of the difference in the makeup of the corpora.Footnote ³

Figure 3. Distribution of the different uses of literally across time; the top figure is based on COHA, the bottom on COCA

All this supports Claridge's (Reference Claridge2011: 109) observation that hyperbolic literally is not common in the corpus data, but goes against McCarthy & Carter's (Reference McCarthy and Carter2004) finding that 91% of the instances of literally are hyperbolic. On the whole, the data may indicate that literally is undergoing a very slow process of semantic change. The very small number of the nonliteral examples shows that the lexeme has not yet been entirely stripped of its propositional meaning. The combination of high-frequency dual use instances and the perception that these uses can be neatly distinguished from the intensifying ones may explain the negative treatment of literally in the usage guides, and the exaggerated perception of the frequency of the instances exhibiting nonliteral use.

The analysis has also revealed that the interpretation of certain instances may be pragmatically conditioned, and that such cases are therefore ambiguous. In the instances of extreme use, for example, the expression modified by literally may not straightforwardly denote an extreme or doubtful case, though the speaker may have intended it as such, and the very use of literally thus signals a potentially extreme reading. This creates a problem of circularity in determining extreme cases, because the use of literally is itself often a clue that the speaker aims to signal that something is surprising, doubtful or incredible. This may be one reason for the strikingly different findings for the use of literally with hyperbolic expressions reported by McCarthy & Carter (Reference McCarthy and Carter2004) and Claridge (Reference Claridge2011). These problems, however, do not affect the results presented here, because almost all ambiguous instances concern those which I consider to be representative of a single category, i.e. dual use, in the present analysis. A more detailed discussion of this issue is beyond the scope of this paper (but see Kostadinova, Reference Kostadinovaforthcoming; Kostadinova, Reference Kostadinovain preparation).

6. Attitudes to literally

As mentioned in Section 3.3, attitudes to literally were elicited by asking participants what they thought about the word. Elsewhere (Kostadinova, Reference Kostadinova and Tieken–Boon van Ostade2018), I focus on the different attitudes that the participants associated with the use of literally. Here, I will be concerned specifically with whether or not the participants accept these newer uses of literally. It seems that, roughly speaking, the comments reflect the prescriptive − descriptive dichotomy. Some of the participants would not accept the word and the variation in its use, usually supporting this view by drawing on arguments evocative of the doctrine of correctness (Finegan, Reference Finegan1980: 10).

1. 10. That word is awful. I try not to use it. People don't use it correctly. (T, f, 35)

2. 11. Language does evolve, and that is the eternal truth. If it changes for a reason other than dumbing down the language, I'm fine with that. But if it's literally just, ‘oh, everyone uses it and everyone uses it incorrectly’, it shouldn't be. I don't think Oxford should have added the second definition of literally to include figurative use of the word [ … ] It's an incorrect use. (J, m, 19)

These comments show that some speakers express attitudes that are very much in line with prescriptive notions of correctness as being superior to actual language use. In other words, usage should satisfy the established standard of correctness. In the case in (11), the speaker even went so far as to claim that linguistic authorities, such as dictionaries (‘Oxford’ referring most likely to the Oxford English Dictionary), may fail to uphold the standards of correct usage.

On the other hand, there were also participants who, as shown in (12) and (13), either expressed no overt criticism of the use if literally or explicitly argued that the word is acceptable, based on their own observations of how people around them use it, or how they themselves do.

1. 12. I literally do that all the time. I say literally when it's never it's never literal. It's like you want someone to believe you so much that like you wanna explain how you feel so you like use hyperbole. (A, f, 19)

2. 13. You know, I don't use [literally] myself very much and I think people use literally so much that I'm probably deaf to it […] But, everyone does it now so I think that may be one that's getting more accepted and more normalised. (R, f, 32)

Finally, one of the speakers who recognised literally as part of their own language use offered a more detailed account of why he saw literally as acceptable.

1. 14. Sometimes I think the use of the word literally is just funny. I have done it before. A lot of us have. ‘I literally wanted to kill myself!’ The majority of people that I know, especially those I interact with, they use literally; they use it a lot. It's almost funny when they use it and it shows more breadth of languages because literally is, I mean, to know what literally means – it's actually a word that I don't think a lot of the general population, especially people who do not have interesting grammar, would actually know what it means. (J, m, 26)

In relation to this final comment, it is worth noting that Israel (Reference Israel2002: 424) makes a similar observation, noting that ‘people use the word in this way precisely because they do understand the notion of literal meaning, and they associate it, naturally enough, with plain speaking and honest expression’. This comment may be indicative of the development of a certain type of prestige associated with the intensifying and hyperbolic use of literally among younger speakers. While I would tend to describe this as covert prestige (cf. Trudgill, Reference Trudgill1972), as it is associated with a nonstandard form, this is still a tentative observation and more data is needed to ascertain the existence of such prestige, and the extent to which speakers express it overtly in a way exemplified in (14). The development of such prestige, however, coupled with the process of language change itself, may turn out to be critical in the eventual acceptance of the extended uses of literally.