Alex Steer

931 words | ~5 min

I'm looking at the moment for research into the application of linguistics to marketing. Compared to the work on applying neuroscience to marketing (known, imaginatively, as neuromarketing), there doesn't seem to be much. If you've seen any, please let me know.

One I have turned up is:

Zhang, Shi, Bernd H. Schmitt, and Hillary Haley (2003), 'Language and Culture: Linguistic Effects on Consumer Behavior in International Marketing Research,' pp. 228-242 in Handbook of Research in International Marketing, ed. S. C. Jain (Edward Elgar Publishing).

It's written by two marketing/business academics and a Psychology PhD student, and describes two sets of experiments performed on a group of native Chinese speakers and a group of native English speakers (none of them English-Chinese bilingual), to see whether linguistic features present in one of those languages but not the other could affect consumer behaviour. I'll only deal with the first of these experiments here, but it's instructive as a model of how not to do research into language, cognition and behaviour.

The paper starts by drawing on the Sapir-Whorf hypothesis, which is never a good sign. For those not familiar, the hypothesis (in its usual strong form) holds that the language a person speaks affects his/her thought about and understanding of the world, and behaviour in it. This sounds like it might be sensible, but, when you think about it, it isn't. Steven Pinker summarises the arguments against well in The Language Instinct, but in brief the main problem is that if you think the hypothesis is true you have to show that language is affecting thought, not just reflecting it. None of these criticisms of the hypothesis (and they've been going round for decades now) make it into this paper. That, frankly, rings alarm bells.

Now, the experiment. Chinese has a lexical class called the classifier, which doesn't exist in English. Classifiers assign nouns to semantic classes: so, for example, the classifier 'ba' denotes things that can be grasped; the classifier 'tai' is used for electrical and mechanical equipment. The authors take the Sapir-Whorf hypothesis and make a classic prediction: that native Chinese speakers will be more likely to see similarities between nouns that have the same classifier than native English speakers will be to see similarities between the equivalent nouns in English.

There's an obvious and horrible problem with this prediction: even a well-designed experiment that showed such an increased likeliness among Chinese speakers would not be showing that the Chinese language was affecting their thoughts. The most it could show would be that the Chinese language reflects certain cognitive categorizations that Chinese speakers are capable of performing. And, of course, we already know that - if the Chinese language couldn't make semantic classifications, classifiers wouldn't exist in Chinese! The logic is perfectly circular.

Nor does it imply, by the way, that because a language doesn't have classifiers its speakers are incapable of grouping nouns semantically. English doesn't have classifiers, but the authors of this paper have no real difficulty in explaining the meaning of various Chinese classifiers in English to readers who obviously have a knowledge of English! We can see that TVs, radios and computers have similarities without the classifier 'tai' to help us out.

So let's throw the Sapir-Whorf hypothesis out. Is there still the potential for a useful experiment here? Absolutely. It could still show that the classifier system make Chinese speakers more likely to make associations between words in the same class than their English-speaking counterparts. This needn't be language affecting thought in any meaningful sense, but it could be the case that the classifiers cause a priming effect.

However, the experiment is not well designed. The participants were given pairs of nouns and asked to rate their similarity. The finding, unsurprisingly, was that Chinese speakers were more likely to rate words in the same class as similar. If this tells us anything, it tells us simply that people, given two objects whose similarity is asserted every time they are mentioned (by virtue of the classifier), will come to think of those things as similar. But this is simply priming, not language affecting thought. However, no attempt seems to have been made to filter out phonological effects. The very fact that both nouns in a presented pair, when included in full sentences, always have the sound 'ba' right before them, might also cause a speaker to see a connection between them that need have nothing to do with semantics.

The other experiment in the set is rather better designed, by the way, since it uses photographs instead of words, thus limiting the potential for priming. However, the potential is still there, and no attempt is made to account for it or eliminate it, nor is there anything to make a convincing case for the Sapir-Whorf hypothesis. The conclusion - that advertising targeted at native Chinese speakers may be better received if it uses words or images belonging to the same semantic class as whatever it's trying to promote - is interesting, but tells us everything about marketing and nothing much about language.