How Does Our Language Shape The Way We Think?

Seems like its been a while since we last grated our linguistic experts. From How Does Our Language Shape The Way We Think? by Lera Boroditsky, the age-old discussion gets reopened:

Such a priori arguments about whether or not language shapes thought have gone in circles for centuries, with some arguing that it's impossible for language to shape thought and others arguing that it's impossible for language not to shape thought. Recently my group and others have figured out ways to empirically test some of the key questions in this ancient debate, with fascinating results.
Being the Programming Languages weblog, issues surrounding languages in general are somewhat tangential. Unlike the linguists, it is generally accepted that programming language syntax and semantics does have a significant effect on design and construction of programs. But like liguistics, one would be hard pressed to isolate the language from the community (culture). My take would be that a large measure of the benefit of looking at new PLs derives from being exposed to differing communities - not just in learning the details of a language.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Sex, Syntax and Semantics

Some of the technical work underlying this essay was in a really interesting paper called Sex, Syntax, and Semantics by Boroditsky, Lauren A. Schmidt and Webb Phillips. There's a bunch more interesting papers linked to from here.

Obvious, in retrospect

I think the degree to which language shapes thought depends on the degree to which thoughts are specific, and words are specifically selected.

At the same time, mental paradigms can affect the types of constructions we describe phenomena in terms of.

In other words, it's not one way or the other, it's both: there is a complex relationship between words stimulating thoughts and thoughts stimulating words. The former occurs during listening, and the latter occurs during speaking, for example.

My 0.02.

Obligatory Response

I enjoyed reading Boroditsky's piece, and I didn't find anything truly surprising in her results.

As a caution against over-interpreting them though, let me offer an analogy to PLs that would give a hint about how to limit the scope of our understanding of the findings.

Imagine that someone was studying the differences between PLs by accumulating data about the compile times for a given program in various languages. It wouldn't be a surprise to find significant variations in those times, and there might be a significant correlation to the characteristics of the PL. It might well turn out to be an interesting study with some interesting consequences for PL use.

But if someone read that study, and on that basis started making wild conjectures about how the computability class for the two languages was fundamentally different, or that the same algorithm would have a different complexity class in the two PLs, we would begin to suspect that that person had over-generalized the finding to absurdity.

I recommend using a similar standard for interpreting human language latency test results and their kin.

(You got me again, Chris... ;-) )

Turing Complete

Wonders if there is a corollary with natural languages that is parallel with Turing Complete? Carrying that further, would it be fair to say that some thoughts are easier to express and communicate in differing languages. (You can speak Latin in any language). In reading stuff on Sapir-Whorf in the past, it always seemed like they got too stuck on vocabulary (Snow, Colors, etc...). Perhaps it's because I grew up with English, but coming up with new words is as easy as Kwyjibo. Wonders if other languages ain't as fluid in their dictionary revisions.

Anyhow, successful bait. SW hypothesis will persist because it is such an easy sell. :-)

Obligatory Follow-up ;-)

Wonders if there is a corollary with natural languages that is parallel with Turing Complete?

I'm sure I've proposed this before here, but can't find it via search, but something like "Anything that can be thought by a person can be expressed" in a language with this completeness property.

The equivalent of the Church-Turing Thesis would be "Any natural human language can express the same ideas as any other."

Carrying that further, would it be fair to say that some thoughts are easier to express and communicate in differing languages.

In a limited and temporary sense, yes. Some ideas will use fewer words in some languages than others, and things like poetry, puns, and culturally specific references are going to be harder to translate, but this rarely impairs core denotation.

Wonders if other languages ain't as fluid in their dictionary revisions.

Even if for some reason you can't add or borrow a word, you can always repurpose an existing one. Consider the sentence: "Sandy sent me a text". It doesn't use any English words that haven't been around for a long time, but it still can express a relatively new idea.

culture

I am in total agreement with the culture comment. Just look at C# and Java, on the surface very similar languages, but typical usage is miles apart.

In real life people certainly manipulate language to shape other people's thinking (though of course that doesn't mean language shapes thinking). I suppose in good PL's it is the same, the author made bad things hard and good things easy, but there is always the weight of the community that is the real pressure. The 'considered harmful' police are probably the equivalent of political correctness -- it doesn't so much shape the way you think or make you better, more it reminds you that there is a large invisible group of people that apparently feel pain when you say 'swinging penis' or declare a public field.

Sapir-Whorf

There's been some fascinating work in the last ten years that has shown how language does change thought. The classical rejection of the Sapir-Whorf hypothesis is that the language one speaks does not affect the concepts one is able to hold in one's head, and AFAIK this rejection hasn't been overturned, though: the Whorfian results tell us about the kind of thoughts and concepts we find easiest to hold, or are likeliest to formulate.

The PL analogy might be with the role Turing completeness plays in language utility. Any Turing-complete language with a given set of fundamental world-interaction capabilities (I/O, interrupt services, &c) is capable of describing the same programs as any other, modulo performance issues and notions of program equivalence. But this does not mean that all PLs are equally suitable programming tools: PLs internalise certain ideas about how one goes about solving programming problems, and that in turn affects what program designs programmers tend to have.

I might summarise some of the nice Whorfian results I've seen; Boroditsky's article doesn't cite much.

If I may...

...add a touch of gilding to your lilly:

the Whorfian results tell us about the kind of thoughts and concepts we find easiest to hold, or are likeliest to formulate.

I suggest that it would be useful to add "to communicate" to your list.

I have many thoughts throughout the day, most of which are probably fleeting. But if I communicate a thought, e.g. to a co-worker, then that marginally increases the probability of that thought recurring, perhaps because at some point it is communicated back to me.

(Pause for massive scaling to occur, as thoughts are at least as lightweight as Erlang processes. ;-)

So I suggest that I'm somewhat more likely to spend time thinking thoughts that are easy to communicate in my immediate context. In most contexts that involve other people, that means thoughts that map easily to words.

co-evolution?

I'm leaning towards the co-evolution of language and thought - sort of like language is a by product of the effort to communicate thoughts. Somehow, thought seems to be independent of language from one perspective - my 1.5 year old son who can't speak any language yet has a lot of structured things to say .. and this structure is something we've been negotiating continuously.

It is a bit clearer in the programming world than in the real world, but there is some contamination by the belief that programming languages live in the domain of mathematics and logic which results in languages being argued as "right" or "wrong" in the absolute sense.

I tend to see programming languages as forms of expression being negotiated among people (plural) and computers. That approach feels more fluid to me. The negotiation can result in different languages if, say, the computers in use weren't Von Neumann - for example if they were connection machines or quantum computers instead, the negotiated language will have very different characteristics.

I said this on the hn.yc

I said this on the hn.yc threads on this topic, but I'm going to be a complete dick and repost it here as well because I'm sure there are ltu'ers who aren't going to see it there, and because this question is pretty important, I think. Note, this applies only to natural languages, not programming languages. :P

------

Psycholinguistics research has repeatedly shown that the amount of influence is minimal, amounting to, in general, forcing you to pay attention to certain aspects of the world more so when you're describing using language than when you're thinking about things in other ways. E.g., you notice color differences more quickly or more accurately when you have to choose between two words that your language could have, than if you were picking a tee-shirt. Or you're more attentive to left and right orientation if your language has words for left and right when you're talking about something positioned relative to your body, but if you're just trying to navigate somewhere or looking for something, your language doesn't affect your thinking.

This effect is precisely what you'd expect, ofcourse, because you get it for absolutely everything else. If your culinary habits involve avoiding high fructose corn syrup, when you're looking for food you're going to be quicker to notice the presence of HFCS than if you're just talking to someone and they mention the corn industry's current HFCS production rates or whatever.

The generalization: when some cognitive task requires that you pay special attention to something, you will. Language is not special in this regard.
And that is how language influences the way we think.

In ultra-condensed form:

Language primes thought.

Priming

Certainly there is this, and many linguists think that this is all there is to these effects. There may be more, though...

.. the pump

There may be more, though...

Such as...? ;-)

More

This is one of those things that I think is true, but I can't prove. When I have more time, I hope to put together a ref list to give an idea as to why I think there is more. But I can at least throw about a bit of jargon for the time being.

I think that simple features such as what cases a language has, and word order, constrain how focus markers can be used to indicate the topic focus (cf. Wikipedia's Topic-focus articulation article stub; one of the referenced authors, Barbara Partee, is not unknown to LtU [1]). It seems to me that this has effects on how language competence shapes what things we find easy to make sense of, and how we go about solving problems, in a way that scales up to have effects on monolingual culture.

Priming writ large

It seems to me that this has effects on how language competence shapes what things we find easy to make sense of, and how we go about solving problems, in a way that scales up to have effects on monolingual culture.

I think it would be unsurprising to find evidence that unmarked word order and topicalization mechanisms in a given language have an effect similar to those reported by Boroditsky. It's unclear to me that these would be fundamentally different in effect from some kind of priming.

When we start to talk about culture, I think the waters get muddied. My observation based on a lifetime spent in multi-cultural environments is that the cultural biases swamp any structural effects of language.

I think a PL analog of this would be the (I think) uncontroversial statement that it would be easier for Scheme programmer to write the Y combinator than a C programmer. One can argue about whether there are structural features of these languages that make this so, and there may be some, but in the end the difference is cultural. The culture of the C programmer doesn't see the problem as a sensible one, whereas the culture of the Schemer privileges that kind of problem.

Culture

It's unclear to me that these would be fundamentally different in effect from some kind of priming.
It doesn't matter: if my second hypothesis is true, then you can't explain quasi-Whorfian effects purely through priming.

When we start to talk about culture, I think the waters get muddied.
No doubt. Still, sociolinguists do study these things.

cultural biases swamp any structural effects of language.
I'm particularly interested in the question of whether speakers of different languages have differences in which inferences they tend to regard as sound or unsound. This is a part of rhetoric that has some relevance to big culture questions such as: which lies can national leaders get away with?

The Clouds

It doesn't matter: if my second hypothesis is true, then you can't explain quasi-Whorfian effects purely through priming.

Well, I certainly agree that if your hypothesis is true, then your hypothesis is true. ;-)

I still don't have a clear idea of what kind of effects arising from which linguistic structures your hypothesis posits that are clearly distinct from priming.

Part of the barrier is definitional: making some task easy by "bringing some mental machinery ready to hand" is a reasonable definition of what priming is.

Still, sociolinguists do study these things.

The null hypothesis in any modern sociolinguistic literature I've seen is that language use encodes culture rather than the other way around.

I'm particularly interested in the question of whether speakers of different languages have differences in which inferences they tend to regard as sound or unsound.

I can think of one class of "reasoning" that fits this bill perfectly: what we normally call "sophistry". Sophistry usually exploits ambiguities of expression in language to "prove" points. Some linguistic ambiguities are going to be idiosyncratic to a given language, so I would have to grant this might be a class of arguments that match what you are suggesting.

This got me thinking about what sophistry would look like in a PL context. "Everything is an object, except when it's not" comes to mind. ;-)

Plato, Aristotle, Socrates? Morons!

Ignoring the historical bias against working class intellectuals... :-)

This got me thinking about what sophistry would look like in a PL context.

I've always associated sophistry with syllogistic errors - it masquerades as deductive reasoning - with the error being introduced with how the premises are drawn up.

Most PL advocacy seems to employ sophistry.

"Share nothing" sophistry

Most PL advocacy seems to employ sophistry.

I can't disagree with that. ;-)

I saw another possible example that is a little more subtle with David Pollock talking about "Share nothing is fail".

In a nutshell, he seems to be saying that if you take a problem that either is or has traditionally been designed as a full-shared-state problem and try to use a "share nothing" solution you will have problems.

It's pretty hard to disagree with that.

The "sophistry" comes in when he concludes something more general about "share nothing" rather than questioning the scope of his application of the expression.

What priming isn't

Priming is like memoisation: if the concepts involved in mental activity have been primed, it all goes faster, but it doesn't change what judgements are reached. Since my second hypothesis talks about reaching different judgements about such things as how best to go about solving problems and which inferences are valid, ergo ...

Priming prime

Priming is like memoisation: if the concepts involved in mental activity have been primed, it all goes faster, but it doesn't change what judgements are reached.

My definition doesn't agree with yours. There is a whole class of (at least what I would call) priming experiments where the whole point is that differential priming results in a different judgment. These are sometimes mixed with and sometimes separate from the latency studies.

My own take is that this is still a case of "ready to hand" data, which explains my phrasing in a previous post.

It's possible that we simply have a terminological disagreement, but I'm not sure.

Two definitions of priming

The Priming (psychology) Wikipedia article has an unreferenced, indefensibly broad definition of priming: Priming in psychology is where an early stimulus influences response to a later stimulus — so priming encompasses all memory effects.

Corsini Encyclopedia of Psychology and Behavioral Science, entry for subception: priming is where the occurrence of one stimulus may make it easier to perceive a later stimulus that is related to it in some way ... Typically priming is shown when one person responds more quickly to the related word than to the [unrelated] word. This only treats perceptual priming, but can easily be generalised to conceptual priming. Our point in question could put a lot of weight on how we interpret the phrase "easier to perceive".

My memoisation model of priming actually would allow for observable distortions of judgements reached through priming through race conditions.

And there MAY be little

And there MAY be little angels that push the planets in their orbits.

When I try to speak German

I have to pay attention which gender a given noun is assigned to; an issue which vanishes in English. Now noun gender is, as Brooks would say it, accidental and not essential complexity. (In most cases, getting the gender wrong does not obscure the meaning, though in a few, it can have humorous or salacious side effects). But still, it's worth noting, I guess...

English actually does have

English actually does have gender, but only in a very limited mostly semantic sense. Waiter/waitress, actor/actress, etc. And it's on nominals, not elsewhere. That said, paying attention to grammatical gender probably has no affect on thought at all, even in the attention/priming sense, because many noun classes/genders are incredibly arbitrary.

Definition of linguistic gender

And it's on nominals, not elsewhere.

Which is to say there is no grammatical morphology associated with this limited pattern, which is why linguists say English has no gender.

Right, for the most part

Right, for the most part English has no real grammatical gender, and certainly nothing even as simple as what German has, nevermind the dozen or so genders/noun classes of Kujamaat Joola! But English does have what Whorf termed cryptotypes — morphologically invisible word properties that are relatively semantically vacuous, but which influence large scale acceptance of constructions.

Wittgenstein

Wittgenstein talked about this eighty years ago. I wonder this is not commonplace among computer scientists.

Scholars don't listen to each other

Linguists actually like what Wittgenstein said about family resemblance concepts, but for the most part he is ignored because, like most philosophers of language, he wrote a lot about what language is without discovering what facts linguists had discovered about language.

Searle is a distinguished exception to this common trait.

A commonplace of commonplaces

Wittgenstein talked about this eighty years ago.

Wittgenstein talked about a lot of things. So have a lot of other people.

Talking about things and resolving them are not the same thing.

Science?

Oh, you are so much science-minded. But knowledge often comes before scientific investigation.

Philosophy

Oh, you are so much science-minded.

I really don't see how you got that from what I said.

But knowledge often comes before scientific investigation.

Wittgenstein had some intuitions, a whole lot of other philosophers had some intuitions, I have some intuitions. These are all a kind of knowledge.

But what do you do when the intuitions don't agree? And have consistently not agreed for decades?

One solution some people take is to choose one position as authoritative and argue from there, but argument from authority, though a venerable technique, has always struck me as unsatisfactory.

Another way

Other people try to figure these things out themselves to their own satisfaction. Having the time and absence of pressure that allowed me to do this is the thing I value most about my doctoral studies.

The only way

Other people try to figure these things out themselves to their own satisfaction.

I think we agree that this is the only sensible way to approach these problems.

It still doesn't help though when the person you are talking with has done the same but come to a significantly different conclusion. ;-)

Bourgeois computer science

I think we agree that this is the only sensible way to approach these problems.
Well, it's a luxury of the idle stipend-rich.

Theorists of the world unite!

Well, it's a luxury of the idle stipend-rich.

I don't know. Some of us have had to find the time and energy without the stipend.

You just have to be demented in the right kind of way. ;-)