Linguistic commentary from a guy who takes things too literally

Candy Canes

Posted by Neal on December 21, 2007

Potential conflicts for recently married couples, as they determine how Christmas will be celebrated in their new household:

Gifts: Do you open some on Christmas Eve, or do you save them all for Christmas Day?

Christmas Eve: Do you go to a midnight service, or an afternoon one? (Or neither?)

The word candy cane: Do you pronounce it with the stress on candy, or on cane?

My wife and I still have not reached a reconciliation on the last item. My pronunciation: candy cane, with stress on the first word. It’s the same stress pattern you get with compound nouns like Christmas tree and Nativity scene. Her pronunciation: candy cane, with stress on the second word, like what you’d do with pumpkin pie or Christmas Day. I just can’t get used to the stress being on cane, so when she says candy cane, I keep wanting to interpret it as contrastive stress instead of ordinary stress. For example, if she says, “I got some candy canes for the boys’ stockings,” I get the weird feeling that, even though I’ve been there for the whole conversation, I somehow walked in on the middle of a sentence, one that must have started like, “I didn’t get candy corn, candy apples, or candy cigarettes for the boys’ stockings –“.

So why would there be a difference in where people put the stress accent on candy cane, anyway? I finally decided to find out. Ingo Plag and his team have done quite a bit of research on stress placement in English noun-noun compounds, including a very clearly written paper here. Here’s what I found out from this paper:

Conventional wisdom is that compounds in English as a rule have stress on the first element (the “Compound Stress Rule”; see also this post by Heidi Harley). Some go so far as to say this rule is without exception, and thus that compounds with the second element stressed are actually not compounds at all, but true syntactic phrases. The problem with taking that line is that there are minimal pairs such as Elm Street and Elm Avenue that have different stress patterns even though their meaning is almost identical, and to say that one is a compound and the other a phrase strikes me as being stubborn, and even a use of circular argumentation. (“We know it’s not a compound because it’s stressed on the right. Why is it stressed on the right? Because it’s not a compound.”)

After acoustic analysis of American English speakers pronouncing newly created compounds, Plag finds that the Compound Stress Rule is exceptionless, provided we’re talking about argument-head compounds. These are compounds such as bartender, in which the head noun (tender) is derived from a verb, and the other noun (bar) is what would have been that verb’s direct object (which is one kind of what syntacticians call a verbal argument; others include subject and indirect object). Every compound like this is stressed on the left. (Or is it? It occurs to me that self-portrait could make some trouble.)

Candy cane, however, is not an argument-head compound. Instead, it’s a modifier-head compound. The head noun is cane, and candy modifies it by describing what kind of cane it is (i.e. one made of candy). For modifier-head compounds, there is much more variability in stress assignment. This can be seen in the four examples I gave up above:, which are all modifier-head compounds: Christmas tree, Nativity scene, pumpkin pie, Christmas Day.

Some have tried to classify modifier-head compounds into groups according to semantic relations, and say that it’s just a fact that compounds with these particular semantic relations are right-stressed: the made-of relation (pumpkin pie), the temporal relation (Christmas day), and others. And what of the modifier-head compounds that fall into one of these classes and end up being left-stressed anyway, like the temporal-relationship compound summer school? Well, if you’re excluding right-stressed structures from your definition of compound, then you can say that an exception like summer school was once a phrase, but through sheer frequency and familiarity, has become a compound and had its stress shifted accordingly. If you don’t exclude right-stressed structures, then you don’t have that explanation available to you.

In any case, Plag demolishes this argument. All the compounds he has his subjects say are fresh, new, and never heard before, which would mean all the modifier-head compounds in the experiment should be right-stressed. Despite this prediction, subjects pronounced some of the novel modifier-head compounds with stress on the left side. Furthermore, even in already-existing modifier-head compounds, being left-stressed does not correlate with how frequently the compound appears in corpora.

So what does determine whether a modifier-head compound like candy cane is left-stressed or right-stressed? Plag concludes that in addition to the kind of semantic relation denoted by the compound, analogy is at work: If it so happens that there are existing modifier-head compounds with the same head as a novel one, the novel one will be stressed in the same way as the existing ones. For example, if you want to talk about a present you give on the occasion of someone’s brush with death, you’re not going to call it a brush-with-death present when you already have in your lexicon compounds like Christmas present, birthday present, and going away present.

So where does that leave me with the strangely stressed candy cane? First of all, this stress pattern is less surprising on a modifier-head compound like this one than it would be on an argument-head compound. Second, the made-of relation is known to favor right-stress, as in pumpkin pie and (moving beyond foods) cotton shirt. But in that case, why would one speaker (my wife) have the right-stressed candy cane while another (me) has the left-stressed candy cane? Plag doesn’t answer this question; as he admits, this study is only the beginning of the story, and “semantics and analogy interact in complex ways.” Perhaps further progress is made in later works on Plag’s webpage, but I haven’t read them yet.

[Update] We had some extended family over for some early gift-giving today, and I discovered that my wife’s sister pronounces candy cane not like my wife does, but like I do! They grew up in the same household; I wonder what happened?

16 Responses to “Candy Canes”

  1. Viola said

    Okay, Mr. PhD in Linguistics….your brains are going a million miles an hour with this one. I’m going to put forth a couple of simple concepts because I can’t possibly keep up with your vocabulary and education; although your knowledge is VERY much appreciated and great food for thought. 🙂 Could it be location, location, location? I’ve noticed there is a huge difference between left-stressed and right-stressed vocabulary with my huband and I, simply due to the fact that we were raised in two different states. Sometimes his family thinks I’ve landed from Mars (perhaps Venus would be a better planet) because I stress the “t” in Dayton, among other things. I think he’s toying with me half the time because he’s constantly leaving 1/2 of a consonant sound out of words that should definitely be stressed and enunciated, otherwise it is considered mumbling. Throw in the fact that we both have ambidextrous and dislexic tendencies for which we have compensated throughout our lives! It all makes for interesting conversation. Gotta love it! Happy Holidays to you and yours.

  2. Kip said

    I’ve always had trouble hearing which syllable is stressed in a word/phrase. When I try saying the examples you gave, I don’t think I’m stressing any syllable. “Christmas tree” doesn’t sound different in stress than “pumpkin pie” to me.

    This gave me troubles in high school Spanish class, as Spanish (IIRC) has pretty strict rules about which syllable to stress. There would be conversations where I would say a word (say “donde”, meaning “where”), and the teacher would say “no, donde”. “that’s what i just said”. “no you said donDE, not donde”. “ok, donde”. “no, you’re still doing it.”

    (for the record, born & raised & still living in NC)

  3. Viola said

    Some people are actually tone-deaf and have a lot of trouble distinguishing between any kind of stresses in language. It can be difficult in communicating, especially when the tone-deaf person responds with a dead-pan facial expression. Some people interpret that as “the lights are on, but no ones home.” That’s not the case, it’s just a matter of how words are heard. One of my mothers (I have 3) and one of my brothers are tone deaf.
    I’ve found that different dialects and accents in different parts of the country can really throw you for a loop. I’m from Washington State (not pronounced Warshingtone!) and when I joined the military, had quite an adjustment in Ft. Jackson, SC. In basic training, getting ready for in-processing, our job was to clean the latrine. One of the gals with a heavy Southern accent came in screaming, “Who touched my breasts? I swear I just finished with that! Who touched my breasts?” Come to find out, she had just polished the “brass” on the doors and someone smudged it. (I probably smudged it, but I DIDN’T touch her breasts!) Anyway, I figured it out, but was freaked out a little in the process of interpretation. 🙂 Back then, Clinton had just instituted the “Don’t ask, don’t tell policy” and I was literally wondering what kind of military I had joined.

  4. Glen said

    Like you, I put the stress on “candy,” which might lend credence to Viola’s regional explanation since we grew up in the same household. I wonder if our deviation from the “made of” rule, which would put the stress on “cane,” derives from the fact that there are few if any other canes of that type. We don’t have chocolate canes and bubblegum canes, for instance. For that reason, I don’t think of a candy cane as a type of cane in the same way I think of pumpkin pie as a type of pie; I think of candy canes as sui generis. The only other “cane” compound I can think of is sugar cane, but that’s not analogous because sugar cane isn’t *made* of sugar, it *provides* sugar.

  5. Neal said

    You’re right; there is a lot of regional variation, and that is a likely place to look for the difference in my wife’s and my pronunciation. You can also have variation based on age, sex, or other factors. But even if the left-stressing or right-stressing of candy cane is associated with geographic region, there’s still the question of why one population of speakers would do it one way, and another population another way. After all, there are plenty of aspects of the language that the two populations will have in common, so why should there be variation on this particular item? What we’d like to find is some rule or set of rules that describes how compounds are stressed, and then be able to say that in Dialect A, rule X is more influential than rule Y, whereas in Dialect B, it’s the other way around; or in Dialect A, we have rule X, but there is no such rule in Dialect B. And this rule would, we hope, be somewhat more general than just saying, “Rule: Pronounce candy cane with stress on the {left, right} word.”

  6. Viola said

    Wow! So really what you’re getting at is the EVOLUTION of stresses in words in regards to regional dialects as well as other factors. You’re wanting to get to the specifics of this evolution to establish a PATTERN to find a RULE to explain why this evolution occurs. Am I off-base with this? I’d love to discuss this more because it helps me understand how to communicate in a more effective manner. I’m going to be “off the grid” for a few days due to the holidays and hopefully will be able to continue this conversation. HAPPY HOLIDAYS TO ALL!
    P.S. I do know how to spell dyslexia. I even looked it up in the dictionary, but my brain flipped mid-gear between reading it and typing it. Talk about a classic example!

  7. Viola said

    What is sui generis? It’s Latin, I believe. Generis means general or regular, does it not?

  8. Ingeborg S. Nordén said

    Sui generis means “in a class by itself”; I think Neil’s point is that candy canes differ so much from other sweets/decorations that the way people stress “candy cane” probably has nothing to do with contrastive meanings.

  9. Lee said

    Merry Christmas, everyone!

    Actually, work by Diana Deutsch and others on the incidence of perfect pitch in speakers of tonal vs. non-tonal languages seems to suggest that perfect pitch may be learned, though it must be fostered at a very early age (and one’s native language can really help out in the process of developing it). It seems natural to me to also deduce that tone-deafness is a state which doesn’t need to happen – otherwise, with over 1/2 the world speaking tonal languages, there would surely be many people who couldn’t communicate in their own native language! See the two links on perfect pitch studies at the top of (Disclaimer: I’m not affiliated with Dr. Deutsch or her work in any way.)

    As a side note, my wife frequently pronounces T.V. as TEE-vee, as do her siblings. She was born in Oakland, but her family moved to the Central Coast of California when she was around 6. I don’t think I put any stress on T.V. at all, but if I do, I think I say tee-VEE. I was born and raised in Southern California.

  10. Viola said

    Thanks for the Latin information. So….generis means “class” or “classification” and sui is most likely a word that means “by itself” or “single” or “only.” So a comparative contrast study would not be conclusive since candy cane truly is in a class by itself.

    Thanks for the website regarding Dr. Deutsch’s studies. I discovered recently that part of the tone-deafness in our family may be due to a type of auditory dyslexia or sensory integration/processing disorder. Perhaps Dr. Deutsch’s studies will provide insight on music therapy for these types of disorders.

    You’ve opened up a contemplative can of gummy worms with your update! So we can most likely rule out the regional dialect theory since the two sisters are not in cahoots with their pronunciations.

  11. Grig said

    I had a manager who pronounced, “Taco Bell” with emphasis like one would say “Tinkerbell.” In fact, it would sound like, “I am going to get some burritos are Tacabel.”

    It’s have to explain, but most people say “Taco Bell” as either two words, or with the emphasis on the word “Bell.” I tried to explain this to her, but like Eliza Doolittle, she couldn’t hear how she was saying it.

    “That’s what I said, Tacabel.” She’d say.

    “No, say, Taco BELL. Emphasis on ‘bell.’ Otherwise, it implies a bell made from tacos, and not a dinner bell that announces that substandard Mexican food can be had at a very low price.”


    “No no no wrong wrong!”

    “Just yew white, ‘enry ‘iggins, just yew white!”

  12. Lee said


    You’re welcome. I’d be interested to know if you have any success with music therapy – if Neal doesn’t mind, perhaps you can follow up here. (Or you can click on my name above to get to my blog…)

  13. Lee said

    Neal – if you will indulge one more comment about tonal languages, do you (or anyone else reading this thread) know how speakers of tonal languages add the nuances afforded non-tonal language speakers by varying tone or pitch? Is sarcasm communicated purely through body language, for instance? (I was discussing this thread, and my tonal-language tangent, with my wife a week or two ago and, in the course of that discussion, we came up with the above question. I tried to do a little research on the topic, but couldn’t find anything that answered the question.)

  14. Neal said

    Not knowing the answer to this question, I passed it on to Language Log’s Bill Poser, who knows something about tonal languages. His response:

    This question rests on two implicit assumptions:

    (a) that in languages with lexical tone, the lexical tones exhaust the tonal resources of the language, so that using F0 [i.e. pitch, NW] for any other purpose would eliminate lexical information;

    (b) you can’t eliminate lexical information, or at least, lexical tonal information.

    Both assumptions are false. We know that (b) is false because tonal languages have phonological rules that in some contexts eliminate lexical contrasts. We know that (a) is false because it is perfectly possible both to realize additional tones and to manipulate the F0 contours due to lexical tones, e.g. by raising or lowering them. So the idea that a tonal language can’t have pragmatic intonation etc. is not well founded.

    There are good examples of such things as question intonation in tonal languages. In Japanese, for example, although in the formal written language questions are marked by an interrogative particle without any special intonation, in the colloquial language it is common to use no interrogative particle and mark questions intonationally. In Hausa, downstep [i.e., a high tone that nevertheless is not quite as high as the previous high tone, NW] in the final phrase is suspended in questions.

    That said, I don’t know of any good survey of how “pragmatic” information is expressed in tonal or non-tonal languages. It may be that there are differences, with a greater tendancy to use non-intonational means in tonal languages.

    Thanks, Bill!

  15. Lee said

    Neal – thanks for passing on my question, and adding the explanatory notes to Bill’s response – and thanks, Bill, for the answer!

    I guess I did indeed have assumption (a) in the back of my mind – though now it seems silly to have held it! As for (b), does this:

    We know that (b) is false because tonal languages have phonological rules that in some contexts eliminate lexical contrasts.

    mean that the phonological rules of tonal languages sometimes remove the ability for hearers to differentiate between different words?

    Finally, I must admit that I never even thought of asking my parents about this question – I don’t usually think of Japanese as being a tonal language. (They tried to teach me Japanese when I was a child, but I stubbornly refused! *Sigh*) I just checked Wikipedia, and it seems that only the Japanese spoken in Tokyo is considered tonal! How interesting…

  16. Al said

    Are you sure “candy” was unstressed? I think of “CANDY CANE” (equally stressed words) as being a cane made out of candy, spoken as if it were not a compound word but some novel concept. It would be like saying, “He has a candy gnome on his lawn.” When you say “CANDY cane”, you are showing that you know this is a compound word and that “Candy cane” means something different from just “a cane made out of candy”. This is similar to the difference between “LADYbug” and “LADY BUG” (a female bug), or the difference between “CHINESE school” (where many American-born Chinese children go) and “CHINESE SCHOOL” (a school in China).

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: