A wealthy, wise traveler who spoke a dozen languages came to the kingdom. He used his learning and wit to quickly attach himself to the Raj as an advisor. The problem is that so fluent and perfect were all the tongues that the advisor spoke that no one could tell his country of origin, and this concerned the Raj's guards. "What if he is a spy from an enemy kingdom?" The guard arranged meals between the advisor and visitors from a dozen different lands, speaking a dozen languages; during all of them, the advisor conversed with them as if he were a native, and none of the visitors could detect even the hint of an accent. The guards became desperate, sure that their Raj was allowing his court to be infiltrated by enemies. Finally, one guard had an idea. At lunch one day, one of the guards took the teapot from the servant and said "I'll handle this." Instead of pouring the tea the guard dropped the full pot of hot tea in the advisor's lap, who promptly leaped to his feet, cursing in Persian.
What does it mean when you stub your toe and grunt or curse? Even if you form a coherent monosyllable, in strict semantic terms it doesn't mean anything (unless you somehow misapprehended your injury as being causally related to copulation, feces, or a deity). It's "meaningful" in the sense that it means, behaviorally, you are suddenly and surprisingly in pain, but that's not linguistic. These kinds of nonsemantic utterances are problematic for philosophers of language because they have no truth value, yet they're clearly important to our linguistic lives. In fact they form a kind of instinctive, basic core of our ability to produce language.
When we think about language, I think we don't take nearly enough advantage of the other slightly less bright critters in our phylum. A few days ago I was at a shoreline park and noticed something interesting. There was a group of plovers pecking the wet sand at the water's edge, when two comparatively massive geese came lumbering toward them. The plover on the side of the group closest to the goose piped a little squeaking call, and all the plovers turned and flew away. Later on, I saw a single goose approach some plovers, and again one plover made the same call; again they flew away. Had I witnessed the ploverese word for "goose" or "fly away"?
Of course not. I observed the ploverese equivalent of what you say when you stub your toe: it's behavior. There's no real free will or cognition involved in either act, any more than there's free will when you move your arms to help you run. The plover call is a totally non-arbitrary act that can only be said to represent anything (have any meaning) insofar as birds call to each other as a warning - but that's the same kind of meaning that your toe-stubbing curse has.
Clearly humans were not always the paragon of animals, and there clearly must have been a time when our ancestors were limited only to non-semantic set-in-stone behavioral vocalizations, and that's why we often look at chimps. In one instance, researchers reported that when a new food (grapes) were introduced, the chimps began to make a different "excited" sound at feeding time. Could this be the chimp word for "grape", at least in that lab? Or are the chimps just excited in a different way, anticipating a different taste, of grapes? Is there a difference?
This is my central thesis, that language developed as, and remains at base, an expression of states of the the central nervous system. It is mostly a description of what's going on inside, not what's going on outside. Of course, in any organism that wants to get its DNA into the next generation, there will be some connection between the outside world, and the organism's internal state (which produces observable behaviors, including language) - but that connection can never be perfect.
This statement attacks the unstated assumption that the primitive content of language is semantic content. That is, that in its most basic form, language began as "grape", not as some excited (but nonsemantic) hooting about getting a certain kind of food. Perhaps the better way to look at language is as a set of behaviors reflecting internal states - vocalizations indicating the fear or hunger or aggression of the organism, which were themselves responses to the outside world. As nervous systems become more complex, the internal states of those nervous systems were more and more able to discriminate finer slices of objects and events in the external world, and therefore the vocalizations became more complex. Eventually, the ability to retain, process and pass around information would be selected for, and at that point there would be an evolutionary feedback loop. In plovers, the language behavior is extremely non-arbitrary and low-resolution by virtue of being filtered through the bird's simpler, less-networked nervous system. Consequently, there can be no subtle gradations in that call of the goose's size or speed or location or disposition, because the plover has no internal state to reflect all those dimensions (even if it can be aware of its own location or disposition).
This immediately puts several problems on new footing. First and foremost, the relatively late spread of genes that influence language (40,000 years ago) makes more sense when you realize that a complex nervous system with the ability to react more "finely" to the outside world would have to appear first. Certain kinds of basic verbalizations (like exclamations of surprise) become less of a puzzle when their lack of semantic content is excused. It is also less surprising that commands are in most languages the most basic form of verbs. Refocusing on language as a reflection of internal states takes some pressure off the Hegelian conundrum of definitions: when I say "I want pizza", there's no question about what "pizza" is to me, although that variable may correspond to a state in you that's different. In this light it's amazing that words align with things in the real world as well as they do - but it's good enough for government work. It may be objected that this places the truth value of statements entirely inside the subjective world of the speaker, but in principle, you could look at the neuron pathways active during an utterance to see whether that really is what they meant by "pizza" when they said it.