So...this is a really interesting question, and I have some ridiculously obsessive thoughts on the presentation of comics. It involves some light narratology and a discussion of directional reading protocols.
The question is about how we read comics. The obvious assumption (though problematic) is that we read comics like we read lines of text: consider each lexical unit in turn, constructing meaning along the way.
Of course, linguists would argue that reading is much more complicated than this, and there are loads of studies done with eye-tracking technology that shows our eyes dart around the page as we move in a generally top-to-bottom, left-to-right pattern.
We also tend to chunk sets of words together using punctuation and spacing to organize words into relational sets. I'm not a linguists, so I am sure I am botching this explanation. The point here is that we don't read individual units in isolation.
Comics further complicates this because words (or even phrases) don't communicate meaning in the same way as panels. Any panel of any comic is packed with meaning in much different way so that there is no meaningful analogy to be drawn between reading comics and reading text.
So the question, again, is how do we read comics? The right answer would be to approach the question by examining comics on their own terms, looking at how the reader engages with the comic in ways that are unique to the medium.
A few caveats: reading digitally and reading in print are different, and comics translated from print to digital formats complicates this discussion even further. Mark Waid gave a good talk discussing these issues in more detail than I could here:
Also, digital comics and print comics are likely meant to be engaged with differently, and so are created differently. The way that we encounter a comic will definitely change how we read it, even in something as simple as the portrait or landscape orientation.
So, with all of that said, one of the more reliable ways to figure out how we read comics is to make use of eye-tracking software to see where people's eyes go when they engage with the text. A few years back, @zackkruse gave a paper on just such an experiment.
Granted, it was an incomplete experiment, but the researchers found that when presented with a page of comics readers tend to jump around the page in a loosely diagonal fashion, working from the top-left to the bottom-right, but not purely left-to-right, top-to-bottom.
The visual elements of comics will draw the eye towards different parts of the comic, breaking the reading order of panels. Consider this comic by Kellet:
Likely, you took in the whole shape of the comic first, giving special attention to the panel with the couch as it is the most visually arresting panel on the page. You might not have taken in the "whole" meaning of the panel, but you certainly jumped to that first.
Presented individually, this would not have been possible. Would the joke still work? Probably. Would it have worked in the same way? Probably not. Seeing the punch line early allows for the reader to read the first three panels in context of that information (however incomplete)
So, I would answer Kellet's question by saying it doesn't matter what the reader wants, but more the effect that he wants to create. The different presentations will allow for different narrative effects.
This was a reductive discussion, obviously. And there are people who are likely far more adept at answering it (@forddent knows more about web comics than I do; @Nsousanis knows a lot more about visual imagination and thinking; and so on). Still...I find this question fascinating
