Data and method

Corpora and Google Books N-gram Viewer were employed heavily for obtaining the data for the present study. The main corpora used in the current study were the Brown corpus, COCA (Corpus of Contemporary American English) (2014), COHA (Corpus of Historical American English) (2014), BYU-BNC (British National Corpus) (2007), and GloWbE (Global Web-Based English) (2013). In addition, the Google N-gram viewer was used for further insights. In this study, the frequency data from COCA, COHA and GloWbE were given in normalized figures per one million words.

In what follows, some details are given about the electronic resources used for the present study. As the largest freely-available corpus of American English, COCA contains more than 520 million words of text and is comprised of five registers. Containing texts from the 1810s to the 2000s, COHA is the largest corpus of historical American English, consisting of texts from different registers. BNC contains 100 million words of British English text from a wide range of registers. The corpus of GloWbE is a large English corpus collecting international English from the internet, containing about 1.9 billion words of text from twenty different countries. For further information on the corpora used, see https://corpus.byu.edu/.

Using a yearly count of N-grams found in the sources printed between 1500 and 2008 in Google's text corpora in English, Chinese, and other six languages, Google Books N-gram Viewer (Google Books, 2010) is an online search engine, which outputs a graph that depicts the historical changes of frequency for a particular phrase (or word). It is also currently the world's largest corpus and the only corpus that enables resolution at a fine temporal scale (yearly) over a long period of time (Michel et al., Reference Michel, Shen, Aiden, Veres and Gray2011). The developers of the viewer aimed to create a new approach to humanities research, which would make it possible to rigorously study the evolution of culture using distributional, quantitative data on a grand scale (Bohannon, Reference Bohannon2010). Google N-gram viewer thus enables further understanding of the relationship between language and its culture.

The punctuation marks explored in the current study are the common ones: period, comma, colon, semicolon, hyphen, question mark, dash, exclamation mark, parenthesis, apostrophe and slash. The frequency distributions of these punctuation marks can be obtained using the corpora mentioned above.

Patterns in the frequency distributions of punctuations marks

Three corpora (Brown, COCA and BNC) were searched to collect the frequency data of each punctuation mark. The frequency of punctuation marks calculated in the current study is relative to word count in corpora. The frequencies of punctuation marks per million are shown in Figure 1.

Figure 1.

The distribution of frequency for all punctuation marks combined in three corpora

In Figure 1, the x-axis shows the rank according to the punctuation frequency distribution; the y-axis shows the normalized frequency per million. Statistical analysis shows that the frequency patterns in the use of punctuation marks in COCA and BNC follow a growth regression model, while they follow a power lawFootnote ¹ for Brown.Footnote ² On one hand, the growth regression model, also called growth curve model, captures how a particular quantity increases over time. Growth curves are used in statistics to determine the type of growth pattern of the quantity— be it linear, logarithmic, or exponential. Logarithmic regression in growth curves might present a heavy-tailed distributional pattern. On the other hand, the power law distribution captures a phenomenon whereby a small number of occurrences are common, while instances of larger occurrences are rare. The power law distribution can be found in a wide variety of physical, biological, cognitive, social and artificial phenomena (Kello et al., Reference Kello, Brown, Ferrer–i–Cancho, Holden, Linkenkaer–Hansen and Rhodes2010; Clauset, Shalizi & Newman, Reference Clauset, Shalizi and Newman2009).

In theory, nature, animals (including people), and even well-designed machines will naturally choose the path of least effort so as to reach the best result. The two distributions shown in the frequency of punctuation marks are both heavy-tailed distributional patterns. The power law can explain the behavior that humans make the least effort (Kello et al., Reference Kello, Brown, Ferrer–i–Cancho, Holden, Linkenkaer–Hansen and Rhodes2010), and the growth model also helps explain least-effort behavior. In such a sense, there is very little difference between the two models to demonstrate the least-effort principle.

The frequency distribution of punctuation marks in different registers

After identifying the general statistical pattern of frequency distributions for punctuation marks, we can also ask whether the frequency distributions vary in the different registers. For example, academic English may use few exclamation marks so as to avoid displaying subjectivity and personal emotions in addressing serious topics. This section will examine how the frequencies of punctuation marks are distributed across various registers in English.

COCA provides five different registers for inspection. The data are depicted in Figure 2. Figure 2 shows the normalized frequency per a million (y-axis) for different punctuation marks (x-axis).

Figure 2.

The distribution of punctuation marks for different registers in COCA

It is necessary for a statistical test to be taken to check whether there are significant differences among these data. AVONVA is a commonly effective method to check this kind of significance. One-way ANOVA is used to test the null hypothesis that the means of several data within a group are all equal. One-way ANOVA test in Excel shows that if F > F crit, we reject the null hypothesis, indicating that the difference among values is real. This is the case, 116.5 (F) > 2.124 (F crit). Therefore, we reject the null hypothesis, that is to say, in COCA, the five registers display different frequency distributions with regard to punctuation marks.

Firstly, academic writing displays the fewest periodsFootnote ³, question marks and exclamation marks, but the highest figures for parentheses and semicolons. This demonstrates that sentences in the academic register are longer than in other registers. It is also clear that academic English seldom displays questions or exclamatory sentences. In addition, parentheses are used extensively in academic texts for citations, or occasionally for explanations, while they are seldom employed in other registers, probably so as not to interrupt the flow of text.

Secondly, fiction shows the highest frequencies of periods and exclamation marks. Sentences in fiction tend to be short. Exclamation marks are widely employed in fiction so as to show emotions or the atmosphere. Additionally, fiction uses the fewest hyphens.Footnote ⁴ In contrast, newspapers and fictions tend to combine words through hyphenation into new units to express up-to-date ideas and temporary concepts (see in the following paragraphs and sections for further discussions at this point).

Thirdly, question marks and colons are used the most frequently in spoken English (written transcriptions of recorded spoken language). In daily communication, people often use questions or exclamations, which is a sign of an informal genre for pragmatic communications. The use of the colon is a little complicated because its functions cover many aspects such as the annunciatory, explanatory, appositive, parallel, etc. The colon can therefore be widely used when spoken language is transcribed into written form so as to express complex situations. That is why the frequency of colon in transcribed spoken English is much higher than other registers.

The newspaper genre has the highest frequency of apostrophes and hyphens. ‘The ‘kiss and tell’ principle expresses the essence of good journalism: keep it short and simple and tell the story’ (Busà, Reference Busà2014: 96), and the principle indicates that news prefers simple and concise language; hence abbreviations (e.g. ‘Strategic Health Authority’ becomes ‘SHA’), apostrophes and hyphens between words (e.g. some verb phrases need hyphens when they are used as nouns, like check -up, break -in, turn -on) tend to be extensively used. As for hyphens, Journalism BBC News style guide (2018) suggests, ‘ Hyphens are often essential, if the text is to make immediate sense’, and the other reason is that hyphenation is easy to integrate several words into a unit, yielding a new meaning which allows journalists to express concepts in novel and appealing ways.

It is also of interest to look at the frequency of punctuation marks in British English (see Figure 3). BNC provides two more registers, NON-ACAD and MISC-miscellaneous, that are not included in COCA. A contrastive analysis between COCA and BNC is also useful for understanding the differences between American and British English.

Figure 3.

The distribution of punctuation marks for different registers in BNC

One-way ANOVA shows that the difference is quite significant (57.67 (F) > 2.04 (F crit)) in BNC. The frequencies of the punctuation marks included in the analysis in these registers are quite similar to the results obtained for American English. However, British English newspapers display the most colons. Conversely, American newspapers use the most apostrophes and hyphens. British fiction uses the most periods, which indicates that the sentences are shorter than sentences in the American English material. However, apostrophes and quotation marks are mostly used in fiction and in this respect BNC data differ from American English.

Frequency distributions of punctuation marks in different varieties of english

The differences across varieties of English have been widely discussed from the perspectives of lexicon, syntax, semantics and discourse. However, few studies have investigated the differences in the uses of punctuation (The Punctuation Guide [2018] is an exception). Yet there may be differences in the frequency distributions for punctuation marks in different varieties of English. This section will examine the frequency of punctuation marks attested for 20 English-speaking countries and regions (for the varieties included in the study, see Table 2). The data were acquired through GloWbE.

SPSS was used to examine their discrete degrees, which are chiefly measured by the values of skewness. Skewness is a measure of symmetry. A distribution is symmetric if it looks the same to the left and right of the center point.

Shown in Table 1, skewness, as a statistical index, is a fitting measure of the discrete degrees within a group of numbers. A symmetric distribution such as a normal distribution has a skewness of 0. When data are skewed left, values for the skewness are negative; in contrast, positive values for the skewness indicate that data are skewed right. As positive skewness increases, the degree of asymmetry increases. For example, the skewness values for all twenty columns are 0.597 for comma and 0.125 for periods. This indicates that the difference of the frequencies of commas used in twenty countries is much larger than the difference of the frequencies of periods. The values of skewness in Table 1 show that the frequencies obtained for parentheses, exclamation marks, apostrophes, hyphens and commas vary greatly. However, the use of periods and question marks varies the least between varieties of English. It implies that the boundaries of statements and questions in English seem to be agreed more anonymously by English speakers with various backgrounds than the practices of other punctuation marks due to both obeying syntactic rules more rigidly and strictly than other punctuation marks.

Table 1:

Statistical values of punctuation frequency in 20 English-speaking countries

As shown in Table 2, the length of sentences is very similar among the varieties of English. In strong contrast to the use of periods, the varieties of English exhibit great differences in the use of commas probably due to many and varied functions of the comma (Partridge, Reference Partridge1953: 14–41) distinguishes at least twelve different functions for commas).

Table 2:

Frequency Distribution of Punctuation Marks in Twenty Regional Varieties of English

The greatest difference concerns parentheses. Pakistani English shows the most parentheses (6448.4/mil), almost double in number compared to Nigerian English (3577.14/mil). UK English uses the most apostrophes (4585.83/mil), 2.34 times as many as the Canadian variety, which uses the least (1958.05/mil). Interestingly, Singapore English uses the most question marks (5214.09/mil) and exclamation marks (3554.11/mil). In Singapore English, there is an abundance of frequently used sentence-final particles (such as har, hor, leh, lor, meh, siah, wat) to express exclamations and questions (Leimgruber, Reference Leimgruber2013: 84–95), which might be one important factor responsible for the highest frequency of using question marks and exclamation marks in Singapore English.

A diachronic perspective

After the synchronic descriptions, a diachronic perspective will provide an interesting picture of the role played by punctuation marks in the development of English. This section will focus on changes in the frequencies of punctuation marks in the history of English from 1500 to 2008. The data are captured through Google N-gram viewer, as shown in Figure 4.Footnote ⁵

Figure 4.

Changes in the frequencies of punctuation marks in Google N-gram viewer (1500–2008)

In what follows, comments are provided on a number of the punctuation marks included in the analysis.

Period: By way of background to the use of the period in the history of English, Liberman (Reference Liberman2011) once launched an American Presidency Project, showing that mean sentence lengths have been falling since the founding of the republic and have undergone a cumulative drop of roughly 50%. Haussamen (Reference Haussamen1994) found that the printed English sentence had become shorter by comparing the number of words in written sentences from 1600 to the 1980s. Haussamen (Reference Haussamen1994) accordingly suggested that the printed sentence will continue to develop into a similar direction over the next two centuries. More than 100 years ago, Lewis (Reference Lewis1894: 34) concluded that the English sentence had decreased in average length by at least one half in 300 years (prior to the 1890s).

As the number of periods in English has continued to rise, the length of sentence is very likely to have become shorter. These findings and predictions are confirmed by Google N-gram data. The graph on the top left of Figure 4 shows that the percentage of periods has continued to increase steadily over the last three hundred years, rising from 3% to 4.3%. Currently, social networking tools reinforce the tendency to use shorter sentences.

Semicolon: A jagged upward trend can be seen in the use of the semicolon in English, peaking around 1800, and afterwards the semicolon experienced a long, smooth decline. This tendency has also been identified in some earlier studies, such as Bruthlaux (Reference Bruthlaux1995).

Early 17^th-century writers used colons, semicolons, and commas interchangeably.Footnote ⁶ The semicolon prospered before the 19^th century; however, its frequency of use has fallen in the last two hundred years. It seems that currently the use of semicolons is associated with difficult or abstract topics; hence writers tend to decline to use semicolons, as previously described. Nunberg (Reference Nunberg1990), author of The Linguistics of Punctuation, holds the view that the semicolon seems to be reserved nowadays for certain kinds of highbrow and high-middlebrow writing (Kelly, Reference Kelly1999). The data in Figure 4 are consistent with the observation that the frequency of semicolon has declined.

Question mark: Since 1800, the frequency of question marks fluctuated mildly, but it drastically increased after the 1970s. This tendency is highly likely to continue in the future because of the rise of social networking media where question and exclamation marks are used frequently.

Exclamation mark: The frequency of the exclamation mark also kept fluctuating between 0.06% and 0.08% during the 19^th century before continuing to increase between the 18^th century and the mid-19^th century with some fluctuations. However, its frequency decreased in the 20^th century, reaching the lowest point in 1960s. This could indicate that there is a reduction in the number of exclamatory sentences used to express feelings and emotions. However, the last half-century has seen an upward trend in the use of the exclamation mark, most likely because of the wide application in social media.

Apostrophe: The data of the Google corpora show that the climax for the frequency of apostrophe was reached in the year 1712. However, the use made of apostrophes afterwards underwent a dramatic fall up until 1850. Strangely, it seems that the use of the apostrophe has become fashionable again in the last 60 years, especially in newspapers and magazines. The revival of the apostrophe is also likely to be the result of the rise of social networking, which will be supported by more evidence in the latter part of this section.

Dash: The dash referred to here is the so-called m-dash, ‘—ʼ, different from a spaced en dash (the m-dash is twice as long as the en dash), used as a break in a sentence or to set off parenthetical statements. The use of the dash increased after 1750, then reached its peak (about 0.35%) in 1860, but afterwards continued to drop up until the 1950s before starting to fluctuate between 0.25% and 0.275%.

As for other punctuation marks, in Google N-gram viewer, commas are missing as they are used as dividers; the colon is not available, either. Hence the COHA is used to supplement this missing data (see Figure 5).

Figure 5.

The change in the frequencies of the comma, colon, and hyphen in COHA (1800–2000)

The frequency of the comma has slowly dropped in the last two hundred years in American English, but the period evinces a reverse tendency to the comma. The decline of the use of commas might be caused by the fact that their rhetorical function has weakened in the history of English. Partridge (Reference Partridge1953: 14) comments: ‘In modern usage, the comma is used predominantly for the grammar, the construction or syntax, of a sentence; formerly the comma indicated primarily the rhetorical pauses, as quite often, it still does.’ His point will be further discussed in the next section.

Hyphenated expressions are compounds with hyphenation, such as short -term, would -be, decision -making. Hyphenated expressions can express up-to-date ideas and temporary concepts (such as, floor -to -ceiling windows, a back -to -back connection, 1980s -style dancing, industrial -scale organic producers), and the hyphenated use can make a phrase become a word, such as the premodifiers in ‘state -of -the -art article’ and ‘top -of -the -line use’. A lexical-grammatical device of this kind seems to have become popular in contemporary English. For American English, the frequency of hyphenated two-word expressions has increased in the last 200 years, as shown in two graphs (having different perspectives: component numbers vs. which part of speech [POS] the expression as a whole belongs to) at the bottom of Figure 5. The same tendency can also be seen in British English.