Animal signals can be about things in the world, not just broadcasts of a state

The argument

Those who hold that animal communication is very complex begin by distinguishing two things a signal can do. It can express the state of the sender, as a cry of pain does, or it can be about something in the world, as a word is. Proponents argue that the second is well attested in other species: there are animals whose alarm calls differ according to the kind of predator and whose listeners respond differently to each, taking to the trees for one and scanning the ground for another, without having seen the predator themselves. On this account the listener is getting information about the world rather than merely catching the caller's agitation, and that is reference rather than expression. Second, proponents point to structure above the level of the single signal. There are species whose calls are combined into sequences, where the combination carries a meaning the parts do not have separately and where the order matters. There are others whose signals are learned rather than inherited, which differ between neighbouring populations of the same species in the way dialects do, and which young animals acquire imperfectly and practise. Learning, variation and combination are precisely the features that make human language productive, and proponents argue that finding them elsewhere, even in simpler forms, makes the difference one of degree rather than of kind. The position does not usually claim that any animal has a language in the full sense. Its claim is that the gap has been systematically overstated because researchers looked for the features of human speech and, not finding them, filed the rest as noise, and that the more closely signalling systems are examined the more structure turns up in them.

Premises

Counter-arguments

Those who hold that animal communication is simple reply that each feature cited has a thinner explanation available. Different calls for different predators need carry no reference: selection can attach distinct calls to distinct escape responses directly, so the listener is running a reflex rather than learning a fact. Order-sensitive sequences, they argue, are typically short, fixed and few, which is association rather than grammar, and local variation in learned calls affects how a signal sounds rather than what can be said with it. On this account the systems retain no capacity to express anything not already provided for.

Rejecting the premises

[Rejecting P1] Differential responses to different calls are fully explained by selection attaching distinct calls to distinct escape behaviours; no information about the world need pass between sender and listener for the listener to act appropriately. [Rejecting P2] The sequences in question are short, few and fixed, and learned variation changes how a call sounds rather than what can be said with it, so neither establishes productive combination. [Rejecting P3] Reference, combination and learning in the human case are open-ended, and sharing a label with the bounded versions found elsewhere does not make the difference one of degree. [Rejecting C] The conclusion exceeds the premises: they show that signalling is more structured than was once assumed, not that it is very complex.