Phonetic marketing
« previous post | next post »
Several readers have drawn my attention to this passage in an Op-Ed by Daniel Gilbert, "Magic by Numbers", NYT 10/16/2010:
The hand is not the only part of our anatomy that gives certain numbers their magical powers. The tongue does too. Because of the acoustic properties of our vocal apparatus, some words just sound bigger than others. The back vowels (the “u” in buck) sound bigger than the front vowels (the “i” in sis), and the stops (the “b” in buck) sound bigger than the fricatives (the “s” in sis). As it turns out, in well over 100 languages, the words that denote bigness are made with bigger sounds.
The sound a number makes can influence our decisions about it. In a recent study, one group was shown an ad for an ice-cream scoop that was priced at \$7.66, while another was shown an ad for a \$7.22 scoop. The lower price is the better deal, of course, but the higher price (with its silky s’s) makes a smaller sound than the lower price (with its rattling t’s).
And because small sounds usually name small things, shoppers who were offered the scoop at the higher but whispery price of \$7.66 were more likely to buy it than those offered the noisier price of \$7.22 — but only if they’d been asked to say the price aloud.
The article's link takes us to a press release ("Small Sounds, Big Deals: How Do Number Sounds Influence Consumers?", Science Daily, 1/21/2010), which in turn takes us to the original paper: Keith S. Coulter and Robin A. Coulter, "Small Sounds, Big Deals: Phonetic Symbolism Effects in Pricing", Journal of Consumer Research, 37, 2010. I'm happy to say that this seems to be an interesting — even surprising — and persuasive piece of research, and that Gillbert's description of it is essentially correct.
(He got a few things wrong — the participants in the "rehearsal" group were asked to "read to yourself, without speaking out loud", for example — but overall, his summary is fine.)
The experiment in question exposed 201 subjects to
… four comparative price ads for a fictitious brand of ice cream scoop, which could be associated with a range of prices. The ads had the identical headline, copy, and illustration; they differed only with regard to the stated regular and sale prices. Prices were placed side by side (3 inches apart) to facilitate serial rather than columnar digit-by-digit comparison.
The regular prices were held constant (\$10 or \$3), and the sale prices were constructed so as to associated smaller prices with "bigger" sounds:
[T]he English number words for 3, 6, 7, and 8 involve front vowels and fricative consonants and therefore are expected to be associated with perceptions of smallness, whereas the number words for 1 and 2 involve back vowels and stop consonants and are expected to be associated with perceptions of largeness. Thus, we chose 6's and 3's to manipulate small perceptions and 2's to manipulate large perceptions and created two sets of regular-sale price combinations…
Thus the regular price of \$10 was associated with two possible discounted prices, \$7.66 and \$7.22. The regular price of \$3 was paired with two possible discounted prices, \$2.33 and \$2.22. In each case, the (hypothetically) bigger-sounding sale price is actually smaller, while the (hypothetically) smaller-sounding sale price is actually larger.
Participants were randomly assigned to one of eight conditions: 2 (rehearsal: sale price vs. no rehearsal) × 4 (sale price phonemes: front vowels/fricatives [\$7.66, \$2.33] vs. back vowels/stops [\$7.22, \$2.22]). The students were instructed that they would be analyzing a video case study involving a local retail department store chain and, as background for that case study, should first look through a booklet of print ads for products carried by the retailer. Participants viewed each ad for 10 seconds. To ensure verbal encoding, participants in the phonological rehearsal conditions were instructed to "read to yourself, without speaking out loud" any sale prices they encountered. They were further told to rehearse the prices internally to keep them in memory because they would be asked to recall sale (but not regular) prices. […] In the non‐phonological conditions, instructions regarding rehearsal of sale prices were omitted. Immediately after viewing the ad booklet, participants completed a 5‐minute distraction task. They then completed a questionnaire to measure their value assessments, purchase likelihood, perceived price discounts (in percentage terms), level of attention to and recall of prices, and ad involvement (in that order) related to the ice cream scoop.
Leaving out the ANOVAs and so forth, I've put some of the more easily-interpretable results in this table (note that the value assessments and purchase likelihood are on a 7-point Likert scale0:
| Sale Price | Discount | Rehearsal | No Rehearsal | ||||
| Estimated Discount |
Perceived Value |
Purchase Likelihood |
Estimated Discount |
Perceived Value |
Purchase Likelihood |
||
| $7.66 | 23.4% | 28.70% | 4.87 | 4.20 | 23.21% | 3.97 | * |
| $7.22 | 27.8% | 25.93% | 3.73 | 3.10 | 27.33% | 4.67 | * |
| $2.33 | 22.3% | 28.10% | 4.73 | 4.20 | 22.50% | * | * |
| $2.22 | 26% | 24.13% | 3.70 | 3.20 | 25.05% | * | * |
*results not reported, said to be "directionally consistent with expectations (i.e., greater for the higher-discounted \$7.22 and \$2.22 sale prices) but not statistically different".
The authors
… considered several possible confounds. Consistent with our instructions, participants paid greater attention to the sale than to the regular price (M=5.11 vs. M=3.72 ; t(100)=10.61, p<.001). However, one‐way ANOVAs across the four price combinations revealed no significant differences in level of attention to the sale price or regular price and no difference in level of ad involvement (p<.05). Also, we observed more accurate recall of both regular and sale prices in the phonological rehearsal conditions (79%, 87%) than in the non‐rehearsal conditions (53%, 59%). However, in each of the four price rehearsal conditions, we found no significant difference in discount perceptions between participants who recalled the sale price versus those who did not. These results indicate that sale price recall is not a necessary condition for phonetic symbolism manifestation. They may also suggest that price discount assessments were made at (or near) the time of target ad exposure rather than at the time of questionnaire item completion, because one might argue that accurate recall of regular and sale prices should assuage (rather than increase) the distortion of discount estimates. Finally, during the post‐experimental debriefing, no participants guessed the hypothesized linkage between independent and dependent variables, implying that the effects of phonetic symbolism occur below the threshold of conscious awareness.
D said,
October 23, 2010 @ 10:58 am
"As it turns out, in well over 100 languages, the words that denote bigness are made with bigger sounds."
What about the other thousands of languages?
[(myl) I believe that this is an infelicity of expression on Gillbert's part — he should have said (and probably meant) something like "according to the findings of a study of well over 100 languages, the words that denote bigness tend to be made with bigger sounds."
Presumably the other thousands of languages are statistically similar.]
goofy said,
October 23, 2010 @ 11:26 am
I learned that the vowel in "buck" and "one" (/ʌ/) is not a back vowel. Wikipedia says it is but notes that it's not as back as it used to be.
[(myl) In the IPA vowel scheme, /ʌ/ is certainly a back vowel. In most varieties of English, at least the American ones, the vowel in buck (however you spell it phonetically) is acoustically somewhat centralized. For example, in the vowel data from Hillenbrand 1995, the "uh" vowel has mean F2 of 1181 among the adult male speakers, compared with 910 for "oa" (the vowel in boat) and 1802 for "eh" (the vowel in bet).
Given that /ʌ/ is unrounded and a bit "lower" (i.e. higher in F1: uh 621 eh 588, oa 498), this is still pretty back.
But in any case, the cited study contrasts the /u/ of two (or whatever semi-fronted version of that the subjects actually have) with the high front vowels /i/ and /ɪ/.]
Twitter Trackbacks for Language Log » Phonetic marketing [upenn.edu] on Topsy.com said,
October 23, 2010 @ 12:06 pm
[…] Language Log » Phonetic marketing languagelog.ldc.upenn.edu/nll/?p=2729 – view page – cached October 23, 2010 @ 10:37 am · Filed by Mark Liberman under Language and business, Phonetics and phonology Tweets about this link […]
George said,
October 23, 2010 @ 2:21 pm
"in well over 100 languages, the words that denote bigness are made with bigger sounds."
Hmm, I wonder which languages. In Arabic, there is a diminutive process in which a back vowel is substituted for the base vowel. Examples: 'kitaab' (book) > 'kutaib' (booklet); 'tifil' (baby) > 'tufail' (small baby).
[(myl) The claim is a statistical one, not a categorical one. The original reference for the survey is Ultan, Russell, 1978. Size-sound symbolism. In: Greenberg, Joseph H. (Ed.), Universals of Human Language, Volume 2: Phonology. Stanford University Press, Stanford.]
Sili said,
October 23, 2010 @ 3:23 pm
Is that why people thought Sagan said "billions and billions"?
marie-lucie said,
October 23, 2010 @ 4:59 pm
With phonological change, it often happens that over time, words which started with a certain set of consonants and vowels end up with a different set (usually partially different, but occasionally completely different). Also, even from the start, "small" consonants might be partnered with "big" vowels, and vice-versa. In the word big, the "big" consonants overpower the "small" vowel (and they also cause it to lengthen), even though the vowels are usually held responsible for the overall aesthetic tone of a word. In the word small, perhaps the initial "s" compensates for the rest of the word, which is "bigger" according to its vowel (which of course has changed over the centuries).
Similar effects have been noted about "light" and "dark" overtones. But in French, jour 'day' has a "dark" vowel, while nuit 'night' has a "light" one. On the other hand, the Latin ancestor stem noct- had a "dark" vowel), and the ancestor of jour is not Latin dies (which had "light" vowels) but the derived adjective stem diurn-, which had a "dark" vowel.
[(myl) Indeed. Given all that, it's surprising that the cross-linguistic statistical tendencies are apparently fairly strong — I should say that I haven't looked into the evidence in detail, but fairly specific claims have been around for decades without (as far as I know) any systematic attempts to debunk them.]
Paul Kay said,
October 23, 2010 @ 6:23 pm
It has been suggested that "The phonetic generalization that can be made is that the expression of size utilizes speech sounds whose characteristic acoustic frequencies vary inversely with size of the thing designated." That quote is from an interesting and provocative paper by John Ohala entitled Sound Symbolism. The paper gives some background to the research reported here and engages in some fascinating evolutionary speculation. Readers whose interest is piqued about sound symbolism might well start there.
marie-lucie said,
October 23, 2010 @ 8:35 pm
I did not mean to suggest that sound symbolism did not exist, only that it was not universal, in particular because of language change.
Pflaumbaum said,
October 24, 2010 @ 4:18 am
This might not be relevant enough to go here, as it's not about phonology – if not, delete of course. But this notion of sonic 'bigness', 'lightness' etc. reminded me:
Does anyone have any idea why the words for 'high' and 'low' in many languages should be used both of spacial elevation and sonic pitch? There's no obvious connection between the two concepts. People have given me lots of ingenious suggestions but none of them are very convincing.
Leo said,
October 24, 2010 @ 4:44 am
There's no fricative consonant in "eight", but a stop instead – at least for most people.
The Liverpool accent does have a fricative allophone of /t/, which is one of its distinctive features for me. According to Wikipedia, Irish English has it too.
Pflaumbaum said,
October 24, 2010 @ 5:30 am
Isn't that Scouse /t/ an affricate [ts]?
Virtual Linguist said,
October 24, 2010 @ 5:54 am
@ Pflaumbaum: As a Scouser myself, I'd say I hear an /s/ sound, or fricative, in the pronounciation of the /t/ in word-final position, eg 'all right'. Kevin Watson of Lancaster University, who has made a special study of lenition in the Liverpool accent, says that affrication of /t/ is common word-initially.
Leo said,
October 24, 2010 @ 5:59 am
Sound Comparisons has lots of audio files for Scouse, and also transcribes the /t/ allophone as [ θ ̠ ].
Chandra said,
October 24, 2010 @ 1:51 pm
Interesting that 7 is associated with smallness here. I've noted in the past (entirely anecdotally, of course) that 7 is often used in numerical exaggerations – e.g. Ren and Stimpy's "47 beeeeeeeellion dollars!" – and wondered if this were due to its having more syllables than the other single-digit numbers.
SimonMH said,
October 24, 2010 @ 9:07 pm
Paul Kay, thank you very much for the link to the Ohala paper; it ties in beautifully with some of the work I am doing on phonetics and verse.
goofy said,
October 24, 2010 @ 9:08 pm
In Ladefoged's "A Course in Phonetics", /ʌ/ is a central vowel in English, but it's a back vowel in the cardinal vowel scheme. I guess that's where my confusion comes from.
Faldone said,
October 24, 2010 @ 9:52 pm
The size connotation in 47 may have something to do with the old antacid commercial that advertised its product as absorbing 47 times its own weight in excess stomach acid. I tend to use 47 for a generic large number with that commercial in mind.
Frank said,
October 24, 2010 @ 11:39 pm
It seems to me to be completely missing the point to focus on the linguistics of this, but whatever.
The larger implication of this study is that enough "shoppers" are willing to pay more for the exact same item because they like the SOUND of the PRICE better, to not only be statistically significant, but to make up a large enough proportion of them to generalize to us all. WTF???
Perhaps numerical illiteracy would be a more appropriate focus? Or the sociology of "consumers" versus the mentally competent? On the other hand, the study seems to explicitly require one to behave like a stupid consumer…
First look through a booklet of print ads for products carried by the retailer.
View each ad for 10 seconds.
"read to yourself, without speaking out loud" any sale prices
Rehearse the prices internally to keep them in memory
Complete a 5‐minute distraction task.
minus273 said,
October 25, 2010 @ 4:45 am
Y.R.Chao described an experiment with Chinese kids. He says "big" and "small", the English words, to the children and ask them to guess which one means "big" and which one "small". Kids do think that "big" is smaller and "small" is bigger.
Pflaumbaum said,
October 25, 2010 @ 5:24 am
@ Leo – thanks. Some of the transcriptions on that sight seem a bit funny though. For instance they have 'quick' as [kʰwɪk] when the woman's saying [kʰwɪx]. And many Scousers realise that final /k/ as a [h].
David Eddyshaw said,
October 25, 2010 @ 5:57 am
Edward Sapir himself published on sounds and size – there's a paper on it in the collection "Selected Writings in Language, Culture, and Personality" ed David Mandelbaum.
Can't give any details, unfortunately, as I am currently in the middle of the South Atlantic.
Pflaumbaum said,
October 25, 2010 @ 6:28 am
You've done well to give us that much, given your predicament.
David Eddyshaw said,
October 25, 2010 @ 6:47 am
@Pflaumbaum:
I appreciate your concern …
I am glad to say I am nonetheless on dry land.
This may possibly be the first ever LL comment posted from St Helena.
Vive l'empereur!
Pflaumbaum said,
October 25, 2010 @ 7:19 am
Ah, good to hear, David. Let us know about the local accent… I've got a friend from St. Helena and her accent, oddly, sounds much closer to Australian or NZ than South African.
David Eddyshaw said,
October 25, 2010 @ 8:27 am
@Pflaumbaum:
Certainly it sounds Southern Hemispherish; there's the shift of the vowel of 'set' towards 'sit' and 'pat' towards 'pet', and a very striking fronting of 'pair' towards 'peer' , all non-rhotic (sorry, no IPA available!)
'Yes' ends up like 'ee-us.'
When talking among themselves, the local "Saints" are quite hard for me (British native English speaker) to follow; when talking to me they shift to a more acrolectal sort of speech. Nobody seems to have any trouble understanding my British RP. The actual grammatical structure of their speech seems to be much like many UK local dialects, double negatives as standard etc and 'don't' 'won't' used for 3rd singular. If it was ever a creole it must be in the last stages of decreolisation; as far as I know (not far) it never was. The population has its roots in a lot of places: UK, India, W Africa, Indonesia, China … I think most came during the long period when the island was run by the East India Company, but I've not been able to find anything very definitive about it.
It would be interesting for a professional to study, I think; I don't know if any ever has.
Pflaumbaum said,
October 26, 2010 @ 4:46 am
Wow, that pair-peer one does sound unusual. When I say it it sounds South African, but it's not according to http://www.soundcomparisons.com/. Or at least, it's not Jo'burg.
This is the second time I've mentioned this book on here today, but… there's a good couple of chapters set on East India Co.-run Saint Helena in Pynchon's Mason & Dixon. Though unfortunately it's possibly the only passage in the book where we don't get any local dialect. For quite a lot of it he's just going on about the wind…
iching said,
October 26, 2010 @ 8:18 am
Sorry to be a little off-topic since I don't have anything sensible to contribute on the marketing experiment.
The front vowel vs back vowel for small/large certainly works for Hungarian (Magyar): kicsi and nagy (IPA /ˈkiʧi/) and /ˈnɒɟ/ according to Wiktionary). Does it even make sense to ask whether this pattern holds for most world languages (with English being an exception?)
@Pflaumbaum:
Regarding your query about why the words "high" and "low" are used for sound pitch, I am also curious, as at first blush the choice seems arbitrary since, presumably, the use of "high" and "low" in English to denote pitch predates the discovery that "higher"/"lower notes are associated with greater/fewer number of cycles per second. The terms "sharp" and "flat" are also used in music of course, to denote higher and lower pitch. I wonder about other languages and what adjectives they use. I believe French uses "aigu" and "grave" (as in the names of the accents). What about Latin, Greek and Sanskrit?
I have heard (but cannot confirm) that Turkish uses the words for "thin" and "thick" respectively. That would make sense I think, if you consider stringed musical instruments.
Pflaumbaum said,
October 26, 2010 @ 10:19 am
Hmm… I think Greek uses ὀξύς and βαρύς rather than ἄκρος and βαθύς for high and low sounds.
altus in Latin can certainly be used for high-pitched voice as well as spacial height, although I'm not sure how good an example that is, since spatially it also means 'deep'. I think low would be gravis, corresponding to its Gk cognate, rather than profundus. For voice, I'd think magna voce also included some connotation of depth.
As for Sanskrit, we need a much more accomplished linguist than me. I remember a couple of words for 'high' but have no idea of their semantic range. The online dictionary gives a load of words for both high and low but no other meanings.
It's a good point you make. It may be that the correspondence isn't as common as I assumed. A similar question applies to sharp/thick, light/heavy etc. though, I suppose – metaphors are being taken from the visual/physical world to describe sounds (unless the sound meanings predate the others of course).
Interesting that in English 'deep' can apply to colours but 'high' can't really, while 'sharp' is possible but I'm not sure 'thick' is.
Pflaumbaum said,
October 26, 2010 @ 10:25 am
Or 'blunt', rather. "A blunt red"… you might find it in a high-end paint catalogue or a novel, I guess.
Jarek Weckwerth said,
October 26, 2010 @ 4:16 pm
@Pflaumbaum: Wow, that pair-peer one does sound unusual.
It's actually a typical (albeit relatively new) feature of NZ English. See e.g. below (first relevant Google hit; WARNING: pdf).
Word recognition and sound merger: the case of the front centering diphthongs in NZ English (Warren, Rae, Hay)
John Cowan said,
October 26, 2010 @ 4:32 pm
David Eddyshaw: While you're there, is it really true that it's called [sɛntɪliːnə] locally?
David Eddyshaw said,
October 27, 2010 @ 5:43 am
@John Cowan:
Alas, I am not in a position to configure this browser to display the second, fifth, and last characters of your transcription properly.
However, the two "Saints" I've just asked said roughly [seintili:nV] (with V being schwa), i.e. pretty much the same as me.
But as I said above, the locals speak significantly differently when conversing informally among themselves rather than to me, so this elicited answer can't be said to resolve the question. I haven't so far overheard the name of the island in an informal context while eavesdropping, at least not that I remember.
On the topic of dialect affiliation, another very striking feature is the extremly frequent use of 'eh' ([ei] with open e) as a tag question marker. I think this is common in South African English too, but I'm far from experienced in these matters.
John Cowan said,
November 18, 2010 @ 1:21 pm
I guess my real question is, is it true that it's stressed on the third syllable ("leen"), unlike a reference to the saint, which would be stressed on "hel"?