Against semantic externalism

by Neil Rickert

Sometime, I think it was back in the 1990s, I was involved in an online discussion of AI. I casually remarked that meanings are subjective. To my surprise, somebody gave me an argument against that. The basic idea was that language is shared in the community, and therefore meanings must be shared, so could not be subjective.

I remained unpersuaded, so I was pointed to Hilary Putnam’s paper “The meaning of ‘meaning'”, where Putnam argues that “meaning is not in the head.”

This post is about why I disagree. The expression “semantic externalism” is commonly used for the view expressed by Putnam and others, that meaning comes from the community rather than from the individual person.


My own experience suggests that people disagree a lot about meanings. They perhaps believe that they are disagreeing about logic or about evidence, but often the real disagreement is about meanings. That’s where I get the idea that meanings are subjective. As an example, look at arguments about “free will”, where disagreements over whether we have free will often look more like disagreements over what we mean by “free will”.

Of course, it is quite possible — and even likely — that what I mean by “meaning” is not the same as what Putnam means by “meaning”. But, if that is the case, then our own disagreement about the meaning of “meaning” is evidence that meaning is subjective.

Twin Earth

In his argument, Putnam introduces the idea of “Twin Earth”. Here, Twin Earth is a planet much like earth, with similar people. And what’s in the heads of Twin Earthians is said to be the same as what’s in the heads of us earth people. But, on Twin Earth, the liquid that they call “water” actually has a chemical composition of XYZ instead of H_2O. Yet the Twin Earthians talk about it as “water” much as we would for H_2O.

This argument actually takes some liberties. On earth, water (i.e. H_2O) is actually an important constituent of what is in our heads. But since Twin Earth contains only XYZ, that cannot be what is in the heads of the Twin Earthians. So Twin Earth can not be quite the twin that it is made out to be. I’m pretty sure this point has been raised by critics of Putnam’s argument. But many philosophers still find the argument persuasive.

Star Trek transporters

I’ll take my own liberties with my response. I want to imagine that I have been transported to Twin Earth using a Star Trek style transporter. So I imagine that I am sitting at a dining table, sharing the evening with Twin Earthians. But I am feeling thirsty, so I want to ask someone to pass the jug of refreshing liquid.

So I ask “Please pass the water”. But, as I ask that, I say to myself “that isn’t really water, but I will use that word for better communications.” So I am using “water” to reference what is in that jug, while thinking that’s not what “water” really means.

A little history might be useful here. I grew up in Australia, and I moved to the USA as a graduate student (in mathematics). When I first arrived in the USA, I found that there were quite a few words where the American meanings differed from the Australian meanings. So I was often saying to myself “I will use that word for the sake of communication, but that isn’t what the word really means.

By now, I have lived in the USA for long enough that I have adapted to using the American meanings. But I presume that adaptation has changed what is in my head.

Reference vs. meaning

It seems to me that Putnam’s argument is really about reference, rather than about meaning. I can readily grant that reference is a community matter rather than a matter of what is in our heads. But it seems to me that reference and meaning are quite different things.

Logic Chips

I’ll illustrate how this can work with an example based on logic chips. Let’s suppose that we have some logic chips, using CMOS technology with a power supply of 5V. These chips function based on the input voltage signals that they receive.

Let’s suppose that chipA switches at an input of 2.4V, while chipB switches at an input of 2.6V. For chipA, we might say that the meaning of “true” is any input voltage greater than 2.4V, while for chipB, the meaning of “true” is any input voltage greater than 2.6V. So chipA and chipB disagree about the meaning of “true”. However, in practical use in actual circuits, the inputs that they have to deal with are alway near 0V or near 5V. So, in actual use, they agree on which inputs they take to be true. They disagree on meaning, but they agree on reference.

We can describe the logic chips as categorizing their inputs into the categories “true” and “false”. As I see it, reference is to the categories, while meaning has to do with how they categorize. They can disagree in how they categorize, but they operate in a world where they still get the same categories even though they use different ways of categorizing.

Reference is external, depending on the world in which they categorize. Meaning is internal, because it has to do with how they categorize.

2 Responses to “Against semantic externalism”

  1. I find your argument here quite persuasive.
    First on *experience*, I think you are right on the argument over, say, freewill.

    On Twin earth, what if what they have on their earth is not water as understood by us, then it can’t be same component. Water is so called because we have a definite formula for it.

    Liked by 1 person


%d bloggers like this: