Friday night, while celebrating with some of my colleagues who had just finished their prelim exams (Congrats, all!), I engaged once more in a classic debate: geek vs. nerd vs. dork. What word means what? What activities count in which category? And so on. I remember having this debate for the first time in AP Physics in high school, and it pops up on the internets with some frequency (see, for example, 1, or 2). Most of these debates are well-reasoned, and impassioned, invoking the etymology (e.g. nerd was a word coined by Dr. Seuss, geek referred to circus performers, dork to male anatomy) and such, but few reference any data.

Any casual student of Wittgenstein* knows, however, that the meaning of a word is its use in the language. It doesn’t matter where it came from a century ago, it matters what it does – what we do and associate with it. So, let’s bring in some data! In this post, I use google hits as a good proxy for typical usages of the words nerd, geek, and dork and associated phrases. I will test some hypotheses about various activities and how likely they are to be referred to as nerdy, geeky or dorky, and similarly, I will make some arguments about what activities are most associated the terms on their own.

Hypotheses

H1 – The Dork Stands Alone: “Dork” will least often be associated with any other word, and most often stand on its own. That is, references will be to a “dork” not a “computer dork”, “gamer dork”, etc.

H2** – Nerds are Geeks, Geeks are Nerds: Most concepts will have a lot of hits for both nerd and geek if they have a lot for either. The two have become conceptually indistinct.

H3 – Phonemes Matter! : Sounds that clash will be disfavored (“Star Trek Dork” has clashing K sounds). Alliterations will be favored. And so on. If this hypothesis is supported, it could explain why these terms have not settled out culturally into three distinct categories. The three categories being somewhat overlapping allows for deviations from patterns which are phonetically favored, which in turn maintains the fuzziness of the distinctions. A larger test of this hypothesis would have to look at whole sentences, and also require someone who knows something about linguistics (which I do not).

Data and Methods

The data for this post come from Google searches performed on 8/28-29, 2010. Each of the three main terms was searched for a total hit count as a baseline. Then I searched each phrase – e.g. “Computer nerd”, “Computer dork”, “Computer geek” and recorded the number of hits. The main dependent variable for activities is the proportion of hits which are in each of the three categories. I also created a second dependent variable “normalizing” each bigram to control for the fact that dork, geek and nerd are used with different frequencies overall. Theoretically, this second d.v. is less clear in its interpretation but it seemed like a good idea. Maybe if we imagine that people are randomly choosing which of the three words – nerd, dork or geek – to use to describe a person-activity pair, we can model that decision as multinomial with baseline probabilities pNerd, pDork, pGeek, which those things being estimated by the relative frequency of each term (in this study, .345 for Dork, .222 for Nerd, and .433 for Geek). Thus the second d.v. is testing for deviations from randomness, accounting for that behavioral model, while the first d.v. is testing overall, how much is this particular activity thought of as dorky, nerdy or geeky (ignoring the fact that overall one of those thoughts is more common). Whatever floats your boat. Full data are available on request. On second thought, full data are available in this convenient google docs spreadsheet. I’m not bothering to discuss the second version of the d.v., because few of the results change much, especially in relative terms.

Findings

First, I find that H1 is strongly supported. For every search term except 1, dork underperforms its overall usage (check the spreadsheet if you want to verify, or the graphic below). That is, #hit for Dork is approximately 35% of all hits for dork, nerd and geek. But for most terms, “____ dork” appears less than 10%, and for only 1 searched term does it have more than 20% of the results share. The one glaring exception is “Star Wars”. For some reason, “Star Wars Dork” appears to be 35% of Star Wars Dork/Nerd/Geek hits.

Second, in re: the Star Wars Conundrum. Star Wars is 35% Dork, 54% Nerd, and 10% Geek. Star Trek is 1% Dork, 50% Nerd, and 49% Geek. Why is there such a huge disparity in the dork figures? Why are there so few Star Wars geeks? I read the lack of Star Trek Dorks as partial support for H3 – the two “K” sounds in the phrase Star Trek Dork make it even more unpleasant to say than it is to smell.*** The emergence and meaning of the “Star Wars Dork” are an interesting anomaly however, which calls for further research. For example, I might ask some older nerds at pub trivia tomorrow.

Third, math is geeky, but physics is nerdy. I think this attacks some of the gender-based explanations. As far as I know, math and physics are both coded as pretty masculine still (any gender and science experts out there want to weigh in?). Yet the disparity is sharp – the “Geek % – Nerd %” for Math is 33%, for Physics it’s -18%. This rules out some alternative explanations, but does not directly speak to my hypotheses.

Fourth, biology is geeky, bio is nerdy. So, along with various scientific and leisure activities, I also ran some abbreviations as a robustness check. So I ran “anthro” along with “anthropology”, and “bio” along with “biology”. The Geek-Nerd spread for Biology is 22%, for Bio it’s -65%! “Bio Nerd” is clearly a set phrase, and yet Biology as an activity is somewhat more geeky than nerdy. I interpret this linguistic overlap as strong evidence for hypothesis 2 – geek and nerd are more overlapping than distinct. Here’s a table with the relative disparity across various terms for how geeky vs. nerdy they are. See if you can find a pattern!

Last, let me present an incredible graphic portrayal of what you might call Dork-Nerd-Geek Space. Size of the circle represents the log of the total number of hits for a given term (i.e. the sum of hits from “bio nerd” + “bio geek” + “bio dork”). Location of the circle is determined by the relative frequency of dork, nerd or geek.

I thank information designer Evan Hensleigh profusely for making this absurdly awesome graphic. Check out his site and other work. As you can see in the chart, almost all of the results are along the “Nerd-Geek” axis, and have little weight on the dork side (except Star Wars). Also, as you can see on the chart, many terms are in the middle. (Note that the “used independently” circle is just that, “geek”, “nerd” and “dork” searched on their own. It represents the overall proportion of google hits.) These activities are not inherently nerdy or geeky, they are just in between. Note that the colors correspond to some rough categories – science, other academic, games, general, fiction, and music.

Conclusions

While I doubt any single work could hope to resolve the epic nerd vs. geek vs. dork debate, I hope that this post has offered some modest evidence. First, I have shown that the dork truly stands alone. Except for Star Wars. Second, I have shown that most nerdy activities are also geeky and vice versa, and when one concept dominates there is no obvious pattern behind it – e.g. not all sciences are nerdy or geeky, and so on. Third, I have offered a small bit of evidence for the hypothesis that phonetics matter, that some phrases just sound better than others. This hypothesis needs more research to be fully substantiated. If correct, however, this phonetic favoring could explain the pathway by which the fuzziness between geek and nerd was produced and is maintained. “Freaks and Geeks” is a great title, but “Freaks and Nerds” has nothing going for it, yet the two phrases have no large difference in meaning.****

I want to offer one large caveat. These results cannot fully substantiate the hypothesis that geek and nerd have become indistinct. I have shown that very few areas of nerdery are not also areas of geekery, and vice versa, but it remains possible that “computer nerds” and “computer geeks” are somehow different (for example). This claim would require deeper textual analysis or ethnographic fieldwork to check. I am relatively confident that such a detailed analysis would reveal few if any substantial and enduring differences in the categories, but cannot be certain from the data presented here. Indeed, the perennial debate over the boundaries of the categories nerd and geek itself suggests the fuzziness which I have sought to establish here.

* Doesn’t that phrase just ooze pretension? Gotta love it!
** Hint – I looked at some of the data before coming up with this hypothesis. Much like every other quant project.
*** I was raised on a healthy diet of both Star Wars and Star Trek (TOS, TNG, DS9 and Voyager), so I have no horse to back in this particular race. Can’t we all just get along? Also, I thank Evan for this interpretation of the lack of Star Trek Dorks.
**** I thank Max for this point.