For my dissertation, I've been thinking a lot about essential indexicals and related phenomena and wondering if we could come up with an actual argument that would show that they or some other sort of "perspectival" element sare essential for us rather than merely appealing to intuitions gleaned from thought experiments with grocery carts and ripped bags of sugar. Like most things in philosophy, it probably won't be a knock-down, prove-it-once-and-for-all sort of argument, but that's okay - it still needs to be done. Here's my first pass at such an argument as it relates specifically to temporally perspectival (tensed) thoughts (note that it is extremely rough and a lot of the discussion is oversimplified or not spelled - these are just quick notes I typed out on my computer and in need of polishing over the next few months):
Why do we need tensed representations (count representations which pick out a time via indexicals, demonstratives, first-person representation, etc. as tensed for my purposes here)? Assume we only had tenseless representations. To have these, we need to explicitly represent times in all our representations of what is the case at a time (or times). So when it is some time t, we need a representation T represents t and, to get us to act at the right time, T needs to get us to do at t the actions to be done at t. Two things are needed here: T must be about t and T must in ordinary circumstances only and almost always cause the appropriate actions at t (or t+1). Consider these requirements in reverse order.
For the latter to happen, T must either only show up at all at t or only show up in a certain way at t (say, in the right functional “box”). How would a system, however, acquire such a T? Inference cannot fully explain this since T must either be inferred from a tenseless or tensed representation. As a matter of logic, a tenseless representation cannot follow from a tensed one. And as far as following from a tenseless one, that (or those) would be the T needing explanation. To constantly keep track of the time, T must be something like one syntactic “date” in a system of dates coordinated with actual times via some kind of clock-ish system.
It looks like such a clock system is required if we are only going to used tenseless representations to get us to act. But now that we have the beginnings of an explanation of how the system acquires T, we need an explanation of the semantics for T and the system of temporal representations associated with it, governed by the internal clock. Either the semantics for this system is determined via description or it is not. If it is, then one option is that it is through determining one time via description and then the other times by their relations to this one time. The other option is that they are interdefined in some way.
If it is determined via description in the former way then the only way to do this is by describing events that occur at that time (or using a description of that time which describes it as coming before or after to a certain degree the events of another time). This will involve either a purely qualitative description or else some sort of description involving using a rigidly-designating name of an event (or indexicals – but that’s not allowed). If it’s purely qualitative then there’s no guarantee that we’ll pick out a specific time or even the right time (we might get the description wrong or more than one time could fit it). And that’s bad if we want to act at the time represented. If, on the other hand, it uses the mental concept of an event then either we are directly hooked up to the even with reference not being determined by description or the reference is in fact by description. If by the latter, we have the same problem over again. The former, however, seems unlikely – how would we get hooked up in this way with a specific particular event without some indexicalish phenomena going on?
It looks like determining the semantics of the system via determining the reference of one representation via description won’t cut it. How about if it’s more holistic? Here we have an even greater chance of being off since, given that the syntactic times are all interdefined and will presumably involve descriptions of what goes on at some or all of the times being represented. The problems with specificity might be better but the chance of error is increased.
If, however, the reference is not determined via any sort of description then we have the same problem as with referring to events – it looks like this requires the use of some kind of indexicalish activity. Since these seem to be the only options for getting us to act in a timely and appropriate way using tenseless representations alone and they either do not work or end up involving tensed representation, it looks like we really do need tensed representation.
Explanations of why we need other sorts of perspectival representations are going to be similar (perhaps including first-person and phenomenal concepts).
