Here's something I've been thinking about and would love some input on. I'm sure this has been discussed at length before, and I know I'm still in a very early stage of learning, so I'll be equally glad about pointers to older threads, and/or literature.
Looking at Briem's page on spacing, I noticed he employs the "glass of water" analogy to describe correct spacing ("the same volume of water that filled the counters should also fill the gaps between the letters"), which is also how I was taught letterspacing at design school.
A different description, which —being more functionally-oriented than merely formalistically descriptive— intuitively seems to make more sense, is this one that I just came across in Peter Enneson's article in TYPO #13 (page 26): "Perceptual distances between letterforms must be great enough to minimize criterial cue confusions or masking. But not be so large that cross-letter binding or constructive conjunction is inhibited."
So here's the question. Especially compared to definions such as Peter Enneson's quoted above, I'm suspecting the "glass of water" model is overly simplistic – mainly because of the issue of optical scaling, the need of which I'm not sure it can accommodate. My impression from a typesetting/typography perspective is that the need for a relative expansion of inter-letter space (OK, tracking) is much increased at smaller point sizes. Now I know that optimally, setting text at a smaller size means employing a dedicated optically-sized cut, which (inter alia) increases the relative width and visible counterspace of the lettershapes. But it seems that the inter-letter space needs to be adjusted more for smaller sizes than the intra-letter space usually is – as in, the adjustments made to intra- and inter-letter space seem not to be linear.
If this is indeed so, then I'm wondering why. Does the effect of lateral interference increase at smaller sizes? Also, if intra- and inter-letter space do not in fact scale in a linear fashion, how can one talk about glasses of water without factoring in target point size?
Or am I mistaken (and most likely, visually misdirected from working too much with non-optically-sized digital fonts) and in proper optical sizing, the scaling of intra- and inter-letter space actually is parallel/linear, so that the concept of optical scaling does not in fact refute the "glass of water" model?