The unexpected attractiveness of snuck
« previous post | next post »
Continue to follow the Saga of Snuck, I thought that I'd check the relative frequency of snuck and sneaked in the LDC's collection of conversational transcripts, which amount to about 25 million words, mostly collected in 2003. These conversations involve people across all ages, regions, socio-economic levels and amounts of education. The verdict? Basically, sneaked is toast.
Specifically, there were 52 instances of snuck versus 5 instances of sneak. Given the historical trend in written American English noted here, which show "sneaked" and "snuck" reaching roughly equal frequencies in the decade starting in 2000, it seems likely that the conversational counts represent the future of a change in progress.
But I also noticed something else rather surprising. Adding up the counts for sneaked and snuck, we get 57 preterite/past-participle forms of sneak, which is just about as many as the count of 58 for the form sneaking.
This is unusual — other words in the class of locomotion and posture verbs generally seem to have around 3-4 times as many instances of the -ing form as of the -ed form:
| sneaking | 58 | |
| snuck+sneaked | 57 | Ratio 1.02 |
| running | 3025 | |
| ran | 886 | Ratio 3.49 |
| walking | 2423 | |
| walked | 740 | Ratio 3.27 |
| standing | 691 | |
| stood | 196 | Ratio 3.53 |
| crawling | 57 | |
| crawled | 15 | Ratio 3.8 |
| driving | 1754 | |
| drove | 519 | Ratio 3.38 |
I don't know what this means, except that people really seem to like snuck.
[Update — Josef Fruehwald writes:
I thought I'd get some diachronic data on ing/ed ratios, since I thought maybe sneak's strange ratio might be related to it's diachronic increase in frequency. Instead, it looks like the 3ish ing/ed ratio is register effect, and in the sources for the Corpus of Historical American English, the typical ing/ed ratio is about 1 (or 0 in log odds). However, sneak had a much higher ratio prior to 1890 or so.
]
anon said,
June 21, 2010 @ 2:52 pm
What are the frequencies for possible substitutes, e.g., "creeping" and "crept"?
[(myl) "creeping" 21, "crept" 6, ratio 3.5.]
Daniel Smith said,
June 21, 2010 @ 3:06 pm
Isn't "snuck up on" kind of an idiom ("Wow, that really snuck up on us")? Is that affecting your results?
[(myl) No. The count for "snuck up on us" in the same corpus is 1. The count for "snuck up on" is 4 — one "them", one "him", one "us", one "one of the guys".
There are plenty of idioms involving the other words as well: "walk out on", "run away with", "stand up", etc.]
kevink said,
June 21, 2010 @ 3:22 pm
Speaking as a rank amateur, it seems that "run, walk, stand, crawl, drive" would be much more common in describing progressive/continuous actions than the furtive, fleeting – sneak. It's the nature of the action. No?
[(myl) You might be right, though we'd have to look at a much larger sample of verbs to test the hypothesis.
Of course, I don't really think that there's some unusual love for "snuck" warping the cited counts. The COCA corpus has "sneaking" 1310, "sneaked" 787, "snuck" 665, which is a similar (1310/(787+665) = 0.90) roughly equal ratio for the -ing form to the preterite/past participle versions, even though the relationship between "sneaked" and "snuck" is more like 1-to-1 than 1-to-10.]
Kyle Gorman said,
June 21, 2010 @ 3:25 pm
A good metric for skewedness or avoidance of forms, due to Richard Wicentowski (see his 2002 JHU dissertation, p. 110) is:
-log(|C_i| / sum \forall j |C_j not C_i|)
where C_i is an inflectional form of the root (or if you prefer, "lemma") C. the higher the number is, the emptier that cell in the paradigm is. Wicentowksi shows this is a reasonable-well-behaved quantity, and is roughly normal (the logarithm scales back extreme outliers, though note the quality is undefined if the numerator is 0, since log 0 is negatively infinite).
So, using BNC counts (yes, conversation would be a lot better, but you can search them at http://corpus.byu.edu/bnc/), I can compare:
stride, strides, striding, strode: 1795
*stridden (a gap that was discussed on LL a while back by Geoff): 1
-log(|stridden| / |stride, strides, striding, strode|) = 7.4928
w/ "ridden":
-log(|ridden| / |ride, rides, riding, rode|) = 2.778
Or, for the "sneaked/snuck" vs. a root above which also has a syncretic preterite and past participle, "walked":
-log(|sneaked, snuck| / |sneak, sneaks, sneaking|) = 1.029
-log(|walked| / |walk, walks, walking|) = 0.74395
John said,
June 21, 2010 @ 3:45 pm
I like Kevink's idea: sneaking is by its nature hard to detect in the present, so it's noticed with hindsight, hence the prevalence of snuck of sneaking.
anon said,
June 21, 2010 @ 4:22 pm
I'd been wondering similar things to what John and kevink mention. "Snuck" has a forensic sort of feel for me. But if that's the case, should we expect similar ratios for similar kinds of words? "Creeping" and "crept" came to my mind—what others are out there?
Kyle Gorman said,
June 21, 2010 @ 5:03 pm
@anon: there's both "crept", but the confound of "creeped … out" (there may also be people with actual doublets. i don't know.)
To motivate my above post a bit: there are some problems with comparing some of the words in the chart above. A few forms ("ran, "drove") are only preterites, but "sneaked" and "snuck" and all the "-ed" forms can also be the past participle (which granted, isn't terribly frequent). But the denominators for "walk", "stand", "sneak" and "crawl" are artificially inflated, at least in comparison to "ran" and "drove". that is, unless
Even then, it's possible the null hypothesis we're testing against (a constant ratio across roots) just isn't empirically correct. In fact there is some evidence it's not: see this MS from Charles Yang … http://www.ling.upenn.edu/~ycharles/papers/zipfnew.pdf
blahedo said,
June 21, 2010 @ 6:31 pm
I was not even aware of the existence of "sneaked" until grad school (well, not consciously; I'd probably run across it at some point and presumed it to be an error). In my intro linguistics class this past term, I brought up "sneaked" (and "dived") in the context of dialect variation and language change, and my students were pretty sure I was putting them on: every one of them had "snuck" (and "dove") so solidly that they were unaware of the other as even a permitted variant. I'd love to see a sneaked/snuck ratio for a demographically skewed data set—facebook/myspace writing, for instance.
Relatedly, I don't think the weak-to-strong shift is nearly as unusual or unlikely as has been suggested throughout the "snuck" post comments; there are a number of verbs that are historically weak but whose preterite forms I avoid because they don't sound right, which strikes me as (at least potentially) an early phase of the weak-to-strong shift. I wonder if any Latinate-root verbs could become strong?
Kyle Gorman said,
June 21, 2010 @ 7:11 pm
@blahedo: some Old French verbs are now "strong", both zero pasts like cost (cf. ModFr. côte) and quit and ablauting ones like dig/dug/dug (from Fr. diguer) and shortening ones like plead/pled/pled. those are all the ones I can remember.
Goplat said,
June 21, 2010 @ 7:30 pm
People who introduce new strong verbs into English ought to have their asses kicken. Or perhaps kuck.
Alan Gunn said,
June 21, 2010 @ 8:37 pm
I like to think of myself as pretty knowledgeable about and fond of standard English usage as of, say, my grandparents' time, but I wasn't even aware of "sneaked" until my son got dinged for using "snuck" in a grade-school paper. Maybe my problem is that my grandparents, upright people all, didn't do much sneaking. Firefox puts a squiggly red line under "snuck," though.
mgh said,
June 21, 2010 @ 9:19 pm
I had the same thought that kevink noted above. there could also be an influence of how long the action takes to complete (you can walk for longer than you can sneak).
comparison with stealing/stole? catching/caught?
Snuck Sneaks into the Lexeme | cassler.net said,
June 21, 2010 @ 9:37 pm
[…] Liberman of Language Log explains the unexpected attractiveness of the word 'snuck'. I thought that I'd check the relative frequency of snuck and sneaked in the […]
Mark Mandel said,
June 21, 2010 @ 10:22 pm
Specifically, there were 52 instances of snuck versus 5 instances of sneak.
Shouldn't that second mention be "sneaked"?
Adouma said,
June 22, 2010 @ 12:53 am
@blahedo:
Lexicalist's Twitter demographics are telling: for 'snuck' the data is as expected – highest for under-25s, progressively lower for older age groups. On the other hand, 'sneaked' isn't even in the system. Though I don't know if that means that nobody has tweeted the word 'sneaked' before or if it's just a negligibly low number.
Lars said,
June 22, 2010 @ 1:55 am
@Goplat: http://verben.texttheater.de/Englisch — have at them.
WAT a guy said,
June 22, 2010 @ 5:56 am
From The Economist this last week (http://www.economist.com/node/16377083):
"Russia did not, however, rush troops to its aid when Ms Otunbayeva asked for them, as the violence in the south span out of control."
Is "span" standard in British English?
stormboy said,
June 22, 2010 @ 7:27 am
@ WAT a guy said: "Is "span" standard in British English?"
I was about to say yes (it sounds perfectly normal to me – I'm British) but a British colleague tells me it isn't. thefreedictionary (online) lists 'span' as an archaic past tense of 'spin'; the Cambridge Advanced Learner's Dictionary lists it at as alternative UK usage (two quick and easy sources that were to hand).
Terry Collmann said,
June 22, 2010 @ 10:49 am
A quick and dirty search on the Guardian website suggests that "span" as the simple past of "spin" is still used in BrE despite the Concise Oxford also calling it "chiefly archaic", but considerably less frequently than "spun": "span round" 10 examples, "spun round" used solely in the simple past tense, 83 examples; "span out of control" 15 examples, "spun out of control" used solely in the simple past tense, 53 examples. I'm a "span/has spun" man meself, but at 57 I'm also "chiefly archaic".
James Kimbell said,
June 22, 2010 @ 2:15 pm
I can vouch anecdotally for snuck's ubiquity amongst the young. In fact, when I hear someone say 'sneaked,' I don't assume that the person has grown up saying 'sneaked' – I assume that the person is consciously choosing to be use the 'proper' (archaic, almost) version.
Plegmund said,
June 23, 2010 @ 4:45 am
I'd have said 'span' only if it's about thread as in 'when Adam delved and Eve span, who then was the gentleman?' – otherwise 'spun'. Like 'hanged' and 'hung'?
I think 'snuck' has a whimsical, self-consciously vernacular feel that people just like. For those who are aware of it,the fact that it's 'officially' wrong actually contributes to the appeal. (Personally I find it twee, but clearly I'm in the minority).
Anthea Fleming said,
June 23, 2010 @ 6:32 am
I am an Australian of pensionable age. I have adopted 'snuck' as a new preterite of 'sneak' for surreptitious movement 'The cat snuck into the cupboard'. I find it amusing. But, in other equally surreptitious situations, I would definitely say 'sneaked' – e.g. 'The student sneaked a look at the exam-paper'. Also 'I was reading under the desk and Jim sneaked on me!' i.e. he informed on me, or as locally expressed 'he dobbed me in!'
ASG said,
June 25, 2010 @ 3:15 pm
I'd like to second what blahedo said:
I don't think the weak-to-strong shift is nearly as unusual or unlikely as has been suggested throughout the "snuck" post comments; there are a number of verbs that are historically weak but whose preterite forms I avoid because they don't sound right, which strikes me as (at least potentially) an early phase of the weak-to-strong shift.
The main example I can think of is "glided", which sounds awkward and ugly to me. I really, really want the word to be "glid." (However, for some reason I don't feel the same way about "elide" or "pride" [as a verb, viz. "prides herself on"]. I think the semantic overlap between gliding and sliding might be what's throwing me. But it does mean I'll go through some strange contortions to avoid using the verb "glide" in the past tense.)
A lot of my friends invent fake strong verbs as a joke. The other day I had to delay gratification on something for some reason, but a favour a friend did "tode me over." But the fact that we do this at all means that strong verbs might have more of a hold on our minds than we think.
“Snuck” sneaked in « Sentence first said,
October 4, 2010 @ 10:41 am
[…] . . . . However, there's also a clear snuckward trend"; in a follow-up post about the unexpected attractiveness of snuck, he concludes that "basically, sneaked is […]
“Sneaked is Toast” | Centanium said,
August 6, 2012 @ 3:14 pm
[…] Writes Mark Liberman: I thought that I'd check the relative frequency of snuck and sneaked in the LDC's collection of conversational transcripts, which amount to about 25 million words, mostly collected in 2003. These conversations involve people across all ages, regions, socio-economic levels and amounts of education. The verdict? Basically, sneaked is toast. […]