Pie charts and bar graphs
« previous post | next post »
Yesterday's Frazz:
Caulfield's joke illustrates several interesting linguistic points.
(1) First, his choice to compound "pie" with "chart" and "bar" with "graph" is irrelevant to the joke, but it does align with the statistics of usage (and also probably reflects the cartoonist's idea about the assignment that Caulfield was given). Since the opposite associations are also possible, this illustrates what we might call lexicographic crystallization. Here are some counts from Google Scholar:
"pie chart" 238,000 "pie graph" 11,400 "bar chart" 274,000 "bar graph" 466,000
So on the pie-metaphor dimension, "chart" has more than 95% of the head-noun total, while "graph" has less than 5%. In comparison, on the bar-metaphor dimension, "chart" has just 37%, and "graph" has 63%. Neither association is categorical, but both are biased in a way that might develop over time into a categorical distinction. It's possible that the current variation reflects different categorical choices by individual authors or editors.
(2) The basis of the joke is the semantics of the compounds. A "pie chart", in the conventional sense that Mrs. Olsen had in mind, is an image that looks like a pie, not an image that charts pies. That semantic relation is not an intrinsic property of English noun compounds, whose second noun can perfectly well denote an action applied to the preceding one. A "haircut" cuts hair, "birth control" controls births, a "shoulder massage" massages shoulders, etc. But individual compound nouns become associated with a particular type of interpretation, so that a "car park" is a place to park cars, not a garden-like space where automobiles are on display. There's a large literature on this general topic — you could start with the survey in section 2.2 ("Semantic Relations in N° Compound Nouns") of this 1992 paper, or explore more recent works on Google Scholar.
(3) Like other Germanic languages, English generally uses complex nominals where Romance languages have prepositional post-modification (with variation in prepositions that's also worth exploring). Thus German Tortendiagramm and Dutch taartdiagram, compared to Spanish gráfico de torta, Portuguese gráfico de pizza, Italian grafico a torta, French graphique à secteurs (or camembert :-)…).
There's more, but that's enough for this morning.
Chris Button said,
February 17, 2025 @ 9:53 am
"Pie graph" is surely incorrect (hence the low usage) since it does not involve the plotting of points on axes, which defines a "graph" as a type of chart.
Giles said,
February 17, 2025 @ 10:02 am
I asked my Portuguese wife about "gráfico de pizza" and she said that they don't use that — "perhaps it's some crazy Brazilian thing". Apparently in Portugal, the much more sensible "gráfico de queijo" is used.
Mark Liberman said,
February 17, 2025 @ 10:20 am
@Chris Button: ""Pie graph" is surely incorrect (hence the low usage) since it does not involve the plotting of points on axes, which defines a "graph" as a type of chart."
Wrong. The OED's first entry for graph has the gloss
A kind of symbolic diagram (used in Chemistry, Mathematics, etc.) in which a system of connections is expressed by spots or circles, some pairs of which are colligated by one or more lines. Also, occasionally the system expressed by one of these diagrams. In abstract terms: A finite, non-empty set of elements together with a set (empty or non-empty) of unordered pairs of these elements.
This is also the sense that underlies terms such as "knowledge graph". And see also the Wikipedia entry.
You may be right that the OED's second entry
Algebra. A graphical representation of the locus of a function; the traced curve of an equation. In wider use: A line or curve representing the variation of one quantity with another, each quantity being measured along one of a pair of axes at right angles.
has influenced the crystallization of "pie chart". But the other meaning(s) of "graph" remain lively.
Mark P said,
February 17, 2025 @ 10:49 am
I think Chris Button’s interpretation is probably more characteristic of people whose work involves the kind of graph he’s talking about. His thoughts are exactly what mine were as I read the post. When I think of graphing something, plotting values of y versus x in the x-y plane is what’s in my mind.
Chris Button said,
February 17, 2025 @ 10:53 am
@ Mark Liberman
I think you're misreading the OED entry. It needs plots on axes because "a system of connections is expressed by spots or circles, some pairs of which are colligated by one or more lines." Those circles are dots at plotted points, not pie chart circles.
Mark Liberman said,
February 17, 2025 @ 11:19 am
@Chris Button:
The point is that "graph" is not only, or even originally, a plot of vector-space points on orthogonal quantitative axes — the OED's first entry for graph (and things like "knowledge graphs" and "graph theory") do not involve vector spaces at all, but rather connections among abstract entities.
And more generally, the etymology is a shortening of graphic, from Latin graphicus, Greek γραϕικός, < γραϕή drawing or writing. Your algebraic sense is a (later) particularization.
A "pie graph" is an alternative particularization — which is relatively rare compared to "pie chart", as the OP notes. But it's not excluded by the basic definition of the term.
Chris Button said,
February 17, 2025 @ 11:37 am
@ Mark Liberman
I actually disagree. A bar graph does not need to have quantitative variables on both axes. But it does need variables on both axes for plotting purposes. Those variables can then be "colligated by one or more lines" if desired. To me "pie graph" represents a misuse of the word, which explains why most people don't seem to say it.
Anthony said,
February 17, 2025 @ 11:50 am
Colligated? I like it. See also Judith Levi 1978.
Noam said,
February 17, 2025 @ 11:54 am
I’m with @Chris Button. To me a graph is primarily a visualization of quantitative data as points on axes, although modifications like drawing a rectangle from the point to the x axis (ie bar graph) are allowed. I think at least one axis has to be quantitative, or at least I can’t come up with a counterexample that I’d call a graph. Heat maps, for example (where both axes are typically qualitative or categorical) are not something I’d call a graph.
I’m also fine with the graph theory/knowledge graph meaning, but that’s just a different thing (also known as a network in similar contexts).
But pie graph is right out.
The source from the Greek is somewhat ironic, because in English “graphical” usually means the opposite of written
Quim said,
February 17, 2025 @ 12:13 pm
I am pretty sure I never found "gráfico de torta" in Spanish yet. The Spanish wikipedia offers "gráfico circular", "diagrama sectorial", "gráfica circular" o "gráfica de pastel" (where "pastel" is closest to "torta" of course). Personally I'd use "diagrama de sectores" but that may be Catalan influence (in which language we also use "gràfic de formatgets" as in Portuguese (and French Camembert).
Mark Liberman said,
February 17, 2025 @ 12:15 pm
@Noam: "But pie graph is right out."
Peeve duly noted.
I was starting to worry that we'd run out of prescriptivist peeving to document — and there are definitely fewer columnists promoting their personal metalinguistic agendas — but maybe the scholarly community will fill the gap.
Quim said,
February 17, 2025 @ 12:20 pm
By the way, Romance languages have two nouns (grafo-graf-graphe/grafico-gràfic-graphique) which desambiguate two possible English meanings. I think the etymology must be back and forth Latin graphicus -> romance graphique -> English graph -> romance graf.
Quim said,
February 17, 2025 @ 12:23 pm
@Noam: No need for it to be quantitative. In math, every map X->Y has an associated graph, which is a subset of XxY. This is purely set theoretic, doesn't even need to be drawn. But one can "always" represent X on an axis and Y on another.
Philip Taylor said,
February 17, 2025 @ 12:24 pm
"I was starting to worry that we'd run out of prescriptivist peeving" — is that "[…] we had run out of prescriptivist peeving" or "[…] we would run out of prescriptivist peeving", Mark ?
RfP said,
February 17, 2025 @ 2:00 pm
I hate to step into this minefield…
But isn't what we're seeing here an OED entry that prescribes how to interpret or use a term, as opposed to the description of how that term is used in day-to-day practice by people who actually use it?
Or, for that matter, how it's interpreted by others, like myself, who have a somewhat-more-than-passing acquaintance with its proper usage, and who instantaneously—spontaneously, even—agree with the "peevers"?
Mark Liberman said,
February 17, 2025 @ 2:13 pm
@Philip Taylor: ""I was starting to worry that we'd run out of prescriptivist peeving" — is that "[…] we had run out of prescriptivist peeving" or "[…] we would run out of prescriptivist peeving", Mark ?"
Why not both?
David Morris said,
February 17, 2025 @ 2:16 pm
He might have annotated the PIE-chart in proto-Indo-European.
Stephen Goranson said,
February 17, 2025 @ 3:10 pm
In this minefield I tread as a rookie, maybe irrelevantly thinking some thinking precedes words.
If I hear rain on the roof I know it before I put it into words.
Same, sometimes, with seeing cut pies or uneven fences?
Temporarily neither descriptive nor prescriptive?
In any case, interesting debate.
Jarek Weckwerth said,
February 17, 2025 @ 3:43 pm
@Stephen Goranson: Very poetic!
JPL said,
February 17, 2025 @ 4:32 pm
@Philip Taylor: ""I was starting to worry that we'd run out of prescriptivist peeving" — is that "[…] we had run out of prescriptivist peeving" or "[…] we would run out of prescriptivist peeving", Mark ?" Mark: "Why not both?"
That's what I call "doing double duty": "we'd" here would be doing double duty. It's an interesting phenomenon that I see every once in a while. The possibility would be fulfilled here if you were to go on to develop the idea of the original sentence further, both that we had come to the end of all the previous favourite peeves and wrt reasons why we are likely not to see any peeves in the future, distinct thoughts both taking that single sentence as a topic sentence, making use of both prongs of the ambiguity.
bks said,
February 17, 2025 @ 5:16 pm
A pie chart is a type of circle graph:
https://en.wikipedia.org/wiki/Circle_graph
Jonathan Smith said,
February 17, 2025 @ 9:31 pm
self-referential has more potential — pie pie chart , bar bar graph, tree tree, waterfall waterfall chart etc. etc…
Chuck said,
February 18, 2025 @ 6:38 am
I would call the first image a column graph (or chart–take your pick). A bar graph plots the variable value horizontally. Had bit of a mental glitch when I saw the column chart called a bar chart but I got over it. Yes, I have spent too much time plotting data in spreadsheet programs…
Xtifr said,
February 18, 2025 @ 2:48 pm
The discussion about "graph" is reminding me of my favorite quote from Dr. Donald Knuth, creator of TeX and author of The Art of Computer Programming. I don't remember the exact words, but it was something like: "I shall follow the usual convention of books on graph theory by using terminology that is different from other books on graph theory."
Barbara Phillips Long said,
February 19, 2025 @ 3:48 am
@Chuck — I have heard graphs called “bar graphs” when the bars are either horizontal or vertical. I don’t recall encountering the term column graph at all.
If someone told me to “graph the results” I might default to a bar graph, but if I were instructed to “chart the results” I might default to plotting points relative to x and y axes. Of course, I grew up doing such exercises on graph paper, which did not lend itself to creating pie charts. If someone told me to “diagram the results” I would be confused, I think. Venn diagrams are popular these days, but I don’t associate them with statistics even when they convey information (mostly I see them used for humor or sarcasm).
Mark Young said,
February 19, 2025 @ 7:55 am
Data point. I just asked my wife, a professor of psychology, whether she used "pie graphs" in any journal articles she published. I had expected her to clarify it to "pie chart", which is what I'd guessed I would do under similar circumstances. But she didn't bat an eye. Not in journal articles, but in other reports she prepares. She had a momentarily negative reaction to the term "pie chart", but then allowed that she'd heard it before. But it sounded "less natural."