1959 books, counted by their punctuation alone. Every chart here measures the marks; the year, genre, author and download rank only sort the books into groups. Point at a dot to see its book.
Over time
Sentences got shorter and semicolons nearly vanished. Each dot is a book at the year it first came out, and the white line is the median for every 25 years. Translations count by the year of the original. The 114 books from before 1600, Homer and Plato among them, are off the left edge.
The white line is the median of every book.
Semicolons: 14.9 semicolons per 1,000 words in 1600–1624, 1.8 in 1950–1974Commas: 91.1 commas per 1,000 words in 1600–1624, 63.1 in 1950–1974Words per sentence: 9.5 words per sentence in 1600–1624, 12.5 in 1950–1974Question marks: 8.8 question marks per 1,000 words in 1600–1624, 5.3 in 1950–1974Exclamation marks: 5.8 exclamation marks per 1,000 words in 1600–1624, 2.3 in 1950–1974Dialogue: 13% of the words inside quotation marks in 1600–1624, 28% in 1950–1974
Genres
Each genre's average fingerprint: every mark's share of all the marks, on the same scale as the fingerprints on the shelf. A book can sit in more than one genre.
Adventure
199 books
,."';:-?!(
Children's
155 books
,."';:-?!(
History and lives
464 books
,."';:-?!(
Horror and gothic
33 books
,."';:-?!(
Mystery and crime
186 books
,."';:-?!(
Novels
518 books
,."';:-?!(
Philosophy and ideas
397 books
,."';:-?!(
Plays
79 books
,."';:-?!(
Poetry
132 books
,."';:-?!(
Romance
143 books
,."';:-?!(
Science fiction and fantasy
150 books
,."';:-?!(
Short stories
170 books
,."';:-?!(
Authors
Can the marks name the author?
Take one book away from its author. Average the fingerprints of each author's other books, and guess whichever author's average is closest to the book. For the 132 authors with at least three different works on the shelf, that guess names the right author for 241 of 708 books, 34%. Guessing at random would be right about 1% of the time.
Each square is one of an author's books, oldest first: pink where the guess named the author. Spread is how far an author's books stray from their own average fingerprint, where 100 is the gap between two books picked at random from the shelf. The lower it is, the more alike their books are.
132 authors, most books first. The list scrolls.
Charles Dickens
The marks named Charles Dickens for 0 of 20 books. The others were taken for Robert E. Howard (7), Alexandre Dumas (3), Fyodor Dostoyevsky (2), Bertrand Russell, Edgar Allan Poe, Homer, Jacob Grimm, John Milton, Jules Verne, Leo Tolstoy and Robert Louis Stevenson. Spread 66.
Their average fingerprint
,."';:-?!(
Over a career: their books in the order they came out (pink), against the median of the shelf's books from within 15 years of each one (dashed). Every author's chart shares one scale.
Too few dated books for a career: it takes four from 1600 on, at least eight years apart.
Pick two books, two authors, or one of each. An author's numbers are the average of their books on the shelf.
Jane Austen
9 books on the shelf
,."';:-?!(
Charles Dickens
20 books on the shelf
,."';:-?!(
Marks per 1,000 words, except words per sentence and dialogue (the share of words inside quotation marks).
Jane Austen
Charles Dickens
Semicolons
12.7
9.4
Commas
73.8
95.4
Colons
1.2
1.8
Dashes and hyphens
9.4
13.1
Exclamation marks
4.1
6.2
Question marks
3.1
5.3
All marks
179.8
239.8
Words per sentence
19.3
17.5
Dialogue
32%
32%
Punctuation twins
Every book has a closest fingerprint on the shelf, named on its own page. These pairs are each other's closest match, by different authors and centuries apart.
Which mark comes next, whatever words sit between, averaged over every book. The apostrophe column is mostly contractions and possessives.
Rows: the mark before. Columns: the mark that comes next. Each number is a percentage of its row.
,
.
"
'
;
:
-
?
!
(
,
43
30
6
5
5
1
6
1
1
1
.
40
25
15
7
4
2
6
1
1
2
"
38
21
16
7
3
1
5
3
3
2
'
40
31
2
8
5
2
6
3
2
2
;
48
25
4
7
6
1
6
1
1
1
:
28
18
29
7
3
2
9
1
1
2
-
30
22
3
4
4
1
33
1
2
1
?
25
13
34
8
2
1
5
7
2
3
!
26
13
29
10
3
1
5
3
9
2
(
31
21
3
4
3
2
5
1
1
29
Marks that touch
The most common clusters of marks with nothing between them, across the whole shelf.
."683,308 times, in 1850 books
,"503,746 times, in 1836 books
?"254,937 times, in 1662 books
!"172,583 times, in 1599 books
.,149,462 times, in 1500 books
),120,645 times, in 1550 books
).95,387 times, in 1300 books
.'88,659 times, in 1469 books
,'79,159 times, in 1405 books
.--68,284 times, in 941 books
,--48,625 times, in 917 books
.)48,459 times, in 1344 books
Popularity
Do the most downloaded books punctuate differently? Barely. Across the 1959 ranked books, the closest link between download rank and any measure is for question marks: a rank correlation of -0.19, where 0 is no link and 1 or -1 a perfect one.
Words per sentence by download rank
Semicolons: no real link (0.03)
Commas: no real link (0.05)
Colons: no real link (0.04)
Dashes and hyphens: no real link (0.01)
Exclamation marks: a weak link (-0.14)
Question marks: a weak link (-0.19)
All marks: no real link (-0.01)
Words per sentence: no real link (0.02)
Dialogue: a weak link (-0.12)
Records
The top three books for each measure, one per author, among books of at least 2,000 words.
Every book, placed by its punctuation alone. Across: how many pauses (commas, semicolons, colons, dashes) a writer uses before each stop. Up: how loud it is, the share of question and exclamation marks. Click or tap a dot to open its book.
The map is wider than your screen. Swipe it sideways to see the rest.