Cognitive Daily

Blogging on Peer-Reviewed ResearchThe Sapir-Whorf Hypothesis — stated in its strongest form — claims that language determines thoughts: if a language doesn’t have a means of expressing a particular idea, then people speaking that language can’t even conceive of that idea. This strong form has long since been rejected: There are plenty of thoughts we can have without having the words to express them.

But there is also little question that the available words do have an important impact on our thoughts. If a language doesn’t have a way to express numbers above 10, for example, then that would probably result in a somewhat different understanding of the world for its speakers compared to speakers of languages with more comprehensive number systems.

It’s difficult to demonstrate these kinds of differences in the real world, though. Perhaps the different conception of the world comes more from growing up in a society that doesn’t value abstract mathematics than from a particular vocabulary limitation.

Recently researchers have figured out an innovative way around this limitation: They invent entirely new things, and new words to describe them. Take a look at this set of objects, used in a recent study led by Gary Lupyan:

i-ccb5ca142177daf8c0fc5c54e7ff4c62-lupyan1.jpg

Each object is different, but they all share similar features. We might be tempted to invent just one new word to describe the entire set of objects. Closer inspection reveals that the eight objects on the left share some features which distinguish them from the objects on the right. If we invent one word, “leebish” to categorize the objects on the left, and another, “grecious,” to categorize those on the right, will people be better at distinguishing between the two types of objects?

Lupyan’s team showed the objects to 48 college students, asking them to imagine that these objects were aliens from a faraway planet. The students themselves were “explorers,” studying these aliens. Their job was to figure out which aliens were approachable, and which should be avoided. The aliens were displayed one at a time on a computer, and an explorer appeared on the screen next to them. The students then used the cursor arrows to move toward or away from the alien. If the student responded correctly, a pleasant bell sounded. A buzz greeted incorrect responses.

Half the students received one extra bit of information: the aliens were labeled as “grecious” or “leebish,” as appropriate, after they made their choice. So these students, they received additional linguistic information to supplement the images — but only after they had made their assessment of the approachability of the alien. Did the labels make a difference? Here are the results:

i-6628b059f8ab111475b241aca5a92e51-lupyan2.gif

The students who saw the labels learned the difference between approachable and unapproachable aliens significantly faster than students who didn’t see the labels — even though the labels gave them no information that wasn’t available in the unlabeled condition. During the testing session, 8 new aliens that hadn’t been seen before (but were clearly members of one of the categories) were introduced. Once again, the students who had seen the labels performed significantly better (no labels were present during the testing session).

Many of the students in the unlabeled group actually reported that they invented their own names to keep track of the two kinds of aliens. A second experiment confirmed the effect using spoken words as labels instead of written words. Interestingly, a third method of labeling, on-screen location, did not produce results significantly different from unlabeled objects.

Here, it seems, we have a clear case of language influencing thoughts. When people have a label for a category of objects, they learn how to identify objects in that category quicker than if they don’t have a label. They’re also better at identifying new objects that they haven’t seen before.

It’s also clear that training can overcome these limitations — so although language does influence thought, its influence isn’t immutable. Just as we can learn a new language, we can learn to have thoughts that aren’t expressible in any language.

Lupyan, G., Rakison, D.H., McClelland, J.L. (2007). Language Is Not Just for Talking: Redundant Labels Facilitate Learning of Novel Categories. Psychological Science, 18(12), 1077-1083. DOI: 10.1111/j.1467-9280.2007.02028.x

Comments

  1. #1 TheBrummell
    January 7, 2008

    Figure 3 in the paper shows the four treatments (unlabelled, written label, auditory label, and location) with error bars. The four treatments gave significantly different results, according to their statistical tests.

    Interesting stuff.

    Is the Sapir-Whorf hypothesis what underlies those rather silly claims that the Aztecs (or whoever) could not physically see Cortez’s ships?

  2. #2 Joel H
    January 7, 2008

    TheBrummel, I have heard those silly claims too but I suspect that what underlies them would be better characterized as an abuse of the Sapir-Whorf hypothesis.

    Was there a significant difference in the size of the effect between the written and spoken labels?

  3. #3 polishalice
    January 7, 2008

    I remember reading some time ago about deaf children in developing countries who do not fully learn language because they are just assumed slow.

    I think this study would be more interesting if they used people with different levels of education and different levels of language development.

    48 college students are probably accustomed to integrating language into their study and memorization schemes (e.g. using mnemonic devices as a memory aid, which is probably why so many of them assigned names to the figures).

    Also, it would be good to record some free association from the college students before the study. The fact that these things are called ‘aliens’ and bear a resemblance to the aliens from the alien movies probably means the students already had a framework for categorizing them worked out.

    The only thing the study could say definitively is that the authoritative labeling system provided by the researchers is better than letting students come up with their own individual labeling system.

  4. #4 The Ethical Atheist
    January 7, 2008

    I had always had thoughts similar to the Sapir-Whorf hypothesis before, but I had never heard of it. Reading about it on wikipedia was pretty interesting. Very interesting article you have here. Thanks.

  5. #5 TheBrummell
    January 7, 2008

    Hi Joel,

    Yes, I think I misspoke: abuse is certainly the correct word for those loopy ideas about invisible sailing ships.

    Sorry, I don’t have access to the paper right now, so I couldn’t tell you about the written vs. audible labels. My rather faulty memory suggests the difference was significant, though.

    I’ll try to remember to get back to you on that tomorrow, when I should have access again (university library logins can be difficult).

  6. #6 Andrea Grant
    January 8, 2008

    Interesting. I first heard of this concept when reading about gender neutral language some years ago, but had never heard a name for it, although I see the wikipedia entry briefly touches on it.

  7. #7 Dave Munger
    January 8, 2008

    The difference between written and audible labels was not significant.

  8. #8 Robert C. Kahlert
    January 8, 2008

    Sorry, I am being dense; the exposition does not make it clear whether the dangerous/not-dangerous feature was correlated with the leebish/grecious feature.

    Put differently, were the unlabeled students effectively learning two features– dangererous +/-, leebish/grecious — that were independent at the same time, while the other students were only learning the dangerous feature? Or is dangerous +/- equivalent to leebish/grecious–and both sets of students were only learning one feature at a time?

  9. #9 Dave Munger
    January 8, 2008

    No, they were only learning the approachable / unapproachable (“dangerous”) distinction. The labels offered no extra information–they were only labels, and only ever given after students had made their categorizations based on appearance.

  10. #10 Utbildning
    January 8, 2008

    It seems intuitive that additional vocabulary is beneficial to support thinking as well as supporting the communication of nuances. Our Inuit friends (f.i. Greenland) as well as the Sámi people (the Laplanders) in the northern part of Sweden have, on average, more words for “Snow” in their languages than what most other languages have. It is of course so that Laplanders can have a more qualified discussion about snow than what any person with a language originating from around the equator can.

    Does fewer words in a language limit thinking and restrain intellectual development? For instance, the language Taki Taki (or Sranan) only has 340 words. This must seriously limit a persons thinking, not to mention communication.

    Luckily Taki Taki, which is a creole language spoken in Suriname, isn’t the only language. Most people in Suriname can speak Dutch and English as well.

  11. #11 Derek James
    January 8, 2008

    Interestingly, a third method of labeling, on-screen location, did not produce results significantly different from unlabeled objects.

    As I was initially reading through the description, I was thinking “Did they just compare linguistic labeling with no labeling, or did they have a condition with non-linguistic labeling?”

    I’d like to know more about how they handled this condition, but I don’t have access to the paper. Did they always put objects from one class on the left and from the other on the right? How much was their attention drawn to the fact that the objects were grouped by location?

    Initially I was thinking they should have a condition where they used colors as labels, with one group of objects being associated with blue, for example, and the other orange. But one could argue that’s still a form of linguistic labeling if they’re using the names of the colors as labels.

    So I think a better condition would be to design labels very much the same way they designed the original stimuli. That is, come up with an arbitrary symbol for each group which does not already have a fixed verbal label, and instruct students that the first category is associated with this symbol while the others are associated with the other symbol.

    I don’t think location is a very good example of a non-verbal label, so I’m skeptical that this condition really contrasts non-verbal vs. verbal marking.

  12. #12 Jim Thomerson
    January 8, 2008

    When I was an undergraduate, I overheard two Geology faculty talking in the hall. One said something like,”There are no words to express my new idea.” The other replied,”Then you’ll just have to make up some new words.” I think anyone familiar with more than one language knows there are things difficult or impossible to translate with exact meaning from one language to the other.

  13. #13 Derek James
    January 8, 2008

    It turns out McClelland has this paper posted on-line here.

    To answer my question about the location condition:

    In the location condition, subjects were told that some aliens lived on one side of the planet, and others lived on the other side. On each trial, after the subject responded (approach/escape) and auditory feedback was given, the alien moved up or down to signal where it “lived.” The motion started 300 ms after response feedback and lasted approximately 400 ms. The trial ended 1,300 ms after the alien stopped moving. Thus, the alien was visible for a longer total time in the location condition compared with the label conditions.

    They also included verification trials to make sure subjects were learning the associations between stimuli and location, which seems sound.

    However, I’m still not convinced that they’ve shown a particular effect of language, per se. The location condition seems to qualitatively different to draw a direct comparison. I’d still like to see a condition with non-verbal labels.

  14. #14 Freiddie
    January 8, 2008

    I think language is kind of stuck to our ordinary life, so we are sometimes forced to label things even if it’s just a custom label, over time.

  15. #15 wolf
    January 9, 2008

    That´s a really interesting study. However, I tend to agree with Derek in #13 above: It´s not really about language in terms of “English” or “Taki Taki” (@utbilding in #10): They used nonsense words, so like Derek I tend to think the effect might not be about verbal labels as such but rather about labels in general (i.e., using non-verbal symbols might have the same effect as nonsense words).

    And of course, since those were nonsense words, the experiment shows nothing at all about how a specific language (e.g. Taki Taki) might restrict intellectual development – after all, people are (obviously) capable of learning new words/labels.

  16. #16 The Vlad
    January 9, 2008

    Well, this is really interesting!

    Yet, after reading the paper, I cannot agree with the assumption that the verbal labels “provide no extra information to the subject.” They must!

    How else could subjects perform better in one condition than another? The information must be there, but buried in the protocol.

    Here is my guess:

    Error feedback (correct/incorrect), on the one hand, tells subjects whether the motor response (approach/avoid) is correct or not. In this sense, feedback is contingent upon the motor response, providing relative information about the category.

    Verbal labels, on the other hand, are provided to the subject independently of the motor response. Labels thereby provide absolute information about the category.

    Combining these two sources of information therefore improves performance. (Exactly how is a nice problem.)

    That’s my two cents, anyway.

  17. #17 Dave Briggs
    January 9, 2008

    I had always had thoughts similar to the Sapir-Whorf hypothesis before, but I had never heard of it. Reading about it on wikipedia was pretty interesting. Very interesting article you have here. Thanks.

    Posted by: The Ethical Atheist | January 7, 2008 8:04 PM

    Yes, I thought it was a very good post too! The difference between the thought life and being able to express it to the outside world is a multifaceted topic!
    Dave Briggs :~)

  18. #18 MyaR
    January 10, 2008

    I cannot let the Inuit (Eskimo) words for snow statement above stand — debunked here (and multiple other times at Language Log). More links here, for the interested: http://itre.cis.upenn.edu/~myl/languagelog/archives/003286.html

    And I’ve never met anyone in linguistics or anthropology who actually believed in the “strong” Sapir-Whorf hypothesis, except those who invented it as a strawman. And none of them had actually read Sapir or Whorf. Well, and a few very naive people who’d never actually taken a linguistics course, so couldn’t read Sapir or Whorf and get anything out of it.

    And thanks for the link to the paper, Derek.

  19. #19 Dave Munger
    January 10, 2008

    MyaR,

    I haven’t actually read Sapir or Whorf, but from what I’ve read *about* them, your points hold true for Sapir but not Whorf. Whorf expressed his the hypothesis in quite stark terms. Here he is, quoted in Pinkers The Language Instinct:

    We cut nature up, organize it into concepts, and ascribe significances as we do, largely because we are parties to an agreement to organize it in this way — an agreement that hold throughout our speech community and is codified in the patterns of our language. The agreement is, of course, an implicit and unstated one, but its terms are absolutely obligatory; we cannot talk at all except by subscribing to the organization and classification of data which the agreement decrees.

    Whorf, of course, was no linguist, and certainly wasn’t a psychologist. He was a fire inspector for the Hartford insurance company.

    As Pinker points out, while serious scientists haven’t ever really had much faith in the hypothesis, it’s had an extremely powerful cultural impact, and I’ve seen lots of English literature scholars defend the strong version quite steadfastly.

  20. #20 MyaR
    January 10, 2008

    we cannot talk at all except by subscribing to the organization and classification of data which the agreement decrees

    Sure. Wittgenstein said essentially the same thing — “Whereof one cannot speak, thereof one must be silent.” (or whatever translation you prefer.) That’s not the same as “you can’t speak about something you don’t have a word for”, which is obvious nonsense.

    Really, there’s no “of course” to “Whorf was not a linguist”. I mean, Wikipedia even says he was! (That’s a joke.) But he did have peer-reviewed linguistics papers published. I think that makes him a linguist by pretty much any reasonable definition, and it is by merit of those papers that his contribution to linguistics should be measured. And he also was a lecturer in anthropology at Yale (under Sapir).

    I agree that people outside linguistics like to defend the strong version. One of the stronger arguments for better and more teaching of linguistics!

    Incidentally, someone from the same general time as Sapir and Whorf who also wrote in a similar vein (and very accessibly) was Dorothy Lee.

  21. #22 Tony Jeremiah
    January 12, 2008

    Hmm,

    I’d have to agree with Derek’s (@11) comment concerning whether this study is showing the impact of a linguistic cue on thought, or, the benefit of having an extra retrieval cue (which just happens to be language-based). To rectify this issue, there would have to be another experiment whereby the comparison group is exposed to an additional retrieval cue (e.g., an arbitrary non-linguistic symbol as Derek suggests). If there is no difference between the non-linguistic and linguistic cue groups, this would suggest that the phenomenon is due to an extra retrieval cue; if there is a difference (in favour of the linguistic cue group), then this would suggest a language-based phenomenon.

    But this type of research would seem somewhat redundant to the extent that it begins to sound a lot like research that uses the Stroop effect as the main paradigm. In that area, there’s been quite a bit about how language impacts non-verbal cognitive mechanisms (i.e., “thought”). Among the more interesting findings, is research with bilinguals showing that it takes a shorter time to name a color when the Stroop stimulus is written in one language and the person is required to name the color of the stimulus in their other language; relative to situation where both the written form and the spoken requirement are the same.

  22. #23 Mikhail
    January 15, 2008

    I agree with the others who don’t think this study is “a clear case of language influencing thoughts”. :) It’s not. As mentioned before, the non-linguistic condition would have to be a color code/association or an associated symbol. Location is a really bad control condition because due to the nature of the task “approach aliens” it becomes relevant – they can be “anywhere”… As far as I know, there isn’t a single study that actually does a good job of separating language influence from visual, conceptual or task related biases. There was a post on this blog about research in Russian about the blue/light blue dissociation – which is bogus research – never been replicated! I think Dave & Greta should be a little more critical of the stuff they publish…

  23. #24 Utbildning
    January 16, 2008

    MyaR. Your Debunk is comparing Inuit languages with English which so happens to be the language with the most words (and synonyms) out of all languages. English is perhaps not the best control or what do you think?

    What about comparing with a language from around the equator?

  24. #25 Mikhail
    January 16, 2008

    1) English is NOT a language with most words
    2) The debunking of the “Eskimo snow words” myth is actually done non-linguistically – people do see and understand the difference between different kinds of snow despite having or not a specific language in their vocabulary. After all, we might not know specialist terms, but that doesn’t prevent us from understanding the concept!
    3) What does a “language from around the equator” have to do with anything? It’s not like they are all similar or have a particular property…

  25. #26 Utbildning
    January 16, 2008

    Mikhail, the equator is a different temperature zone than the arctic. Hence, snow is rare at those latitudes. Even though I am not specialized in this particular area my guess is that languages originating from “around the equator” have fewer words for snow, if any (except, maybe, for languages spoken close to high mountains such as the Kilimanjaro in Tanzania. There they speak Swahili, Chaga etc). Consequently, an individual from that area would be in a similar situation as the respondents in the unlabeled treatment in the article above when facing various kinds of snow, i.e. they have no words (yet) to describe an object.

    I may be wrong of course since this isn’t my cup of tea :-)

  26. #27 kendall
    July 1, 2008

    As Dave says, in post #19 above, Whorf was not a linguist. It is my understanding that his papers were pseudo-linguistic, published mainly in The Fleet Journal and other state-sanctioned publications for the general audience. And as far as I am aware, all of those publications dealt only with the extraordinary ability of Klingons to master all aspects of the English language except for the use of contractions.

The site is currently under maintenance and will be back shortly. New comments have been disabled during this time, please check back soon.