It is surely true that language as a topic is not exhausted by the norms of a community in the abstract, but also the cognitive and linguistic system of the individual idiolect. So it is also surely true that meaning is not “conveyed”, in a sense, but always self-generated. And this model gives us insight. For instance, it gives us a rough-and-ready idea about how semantic shifts take place; why “muscles” now means what it does, and does not mean what it used to (”little mice”).

That having been said, we may find that the difference in effect amounts to nothing. I say a novel word, “Burnarbapus”; I point to something that has certain features — the features (say) of prickliness, green-ness, upwards length, tends to be in a pot or in the ground, cylindrical with little stubby branches, etc. If you then in turn point to a basset hound and say “Burnarbapus”, then I will correct you; while if you point to a cactus, I nod. Something about the external environment has obviously had an impact upon you in this instance — namely, both my presumptions as a language-bearer, and (more importantly) the pattern-generating features of the ostensive object(s). Whether we call this “structural coupling”, or “social interaction”, or whatever, the bottom line is that something we take to be external (the pattern) has had an influence upon a system that we take to be internal (the idiolect). Without any semantic anchor, as provided by experience and thought, there is no meaning at all; we will have committed semanticide.

I don’t dispute your remarks about consentual domains, etc. I sympathize with them (and, if my PORG project is to go anywhere, I *must* sympathize with them). But I am having a hard time understanding how it provides a counterexample to the “Rob’s mom” point. The question is, would a person be right to say that Rob’s friend has misunderstood the meaning of the utterance, if they make such a grievous error as the one in the example? If all this were contingent upon truly *consentual* domains, and we presume that the friend has already acquiesced to the knowledge of Rob in determining the meaning of “Rob’s mom”, then we’d also need to say that the friend has been in error, as if he had broken a contract. And if the domains were not consentual, then they would be miscalibrated, and no utterance like “Rob’s mom has a daughter” could have a relevant cognitive impact on the friend: it would not meet with any apprehension conditions, due to a lack of conversational effort on the part of the friend to engage in the behavior necessary to communicate.

To come full circle, I believe there is a way out of this puzzle without having to concede semantic ground to the “thing theorists”: we simply insist upon the Pragmatics/Semantics distinction, or the distinction between act/utterance and genuine content/speech. We may say that things are only a part of our meanings insofar as they are referred to — and they are only a part of our utterances, not our language itself. The friend will have minimally apprehended the expression “Rob’s mom”, but will not have understood the conversational utterance involved in the above example.

So to the extent that Cosim’s objections to the act/content divide are cogent, the project for understanding language which you and I share will have to be abandoned (apart from any ethical/legal consequences that we also discussed).