Ned was nice enough to point out that the proofs of his response to us are available online. I want to thank him for his engagement but there is a lot I don’t agree with. I want to say something about each section but first I wanted to address his claim that the argument from Inattention Inflation is question begging. He is wrong about that
He says,
Apparently, their argument is this:
- The first-order states were about the same in strength as evidenced by the equal performance on discriminating the gratings;
- But as reflected in the differing visibility judgments, the unattended case was higher in consciousness;
- To explain the higher degree of consciousness in the unattended case we cannot appeal to a first-order difference since there is no such difference (see premise 1). So the only available explanation has to appeal to the higher-order difference in judgments of visibility.
He then agues that the only reason we would have for accepting premise two of the above argument was a prior commitment to the higher-order thought theory, which is clearly question begging.
First I would object to the characterization of our argument. Premise 2 should not say that one case was higher in consciousness but rather that there were phenomenological differences between the two cases. If there is a difference in what it is like for someone when we have reason to think that there is no difference in their first-order states, then we have reason to think that phenomenology is not fully determined by first-order activity. Block seems very confused by this but isn’t there an obvious difference between clearly seeing something presented to you and just catching a quick glimpse of something or other presented near threshold?
I think that ultimately his argument in his reply to Inattentional Inflation (II) is that since we have two models that both predict the same pattern of results we cannot use the pattern of results as evidence for one model over the other. The two models are
- (A) a first-order view where difference in task performance is indicative of no difference in conscious experience and difference in report is indicative of cognitive effects without necessarily effecting phenomenology.
- (B) a higher-order view where difference in task performance is not indicative that conscious experience is the same and difference in report is indicative of an effect on phenomenology.
The question then comes down to which of these two models we should prefer.
In giving our answer to this Block edited a quote from us without indicating that in the text. We say “if a combined increase in the frequency of saying “yes I see the target” and higher visibility ratings is not good evidence that phenomenology has changed, what else can count?” and he quotes us as just saying if “higher visibility rating is not good evidence…” totally ignoring that we explicitly said it is the combination of both that we are replying on. This is misleading!
It is both of these that lead us to think that there really is a difference between the two cases and that leads us to think (B) is the right interpretation. They say they see it more often and also rate it as more visible even though they are not doing a better job of detecting the stimulus. It has nothing to do with the fact that we are willing to defend a higher-order approach to consciousness.
It is too bad that Lau et al do not collect anecdotes from participants but I think just from our ordinary everyday experience we have some cases of inattention inflation. Sometimes as I am sitting at my computer writing something I will think that I saw the little red icon in the right corner of the screen that alerts me to an email in my inbox. Sometimes I will check and it will indeed be there. Other times I check and there is no red marker. But it sure did seem like there was one there just before I looked! The idea is that something like this is going on in the experimental conditions. I predict that if asked subjects would be surprised to find out that (some of) their false alarms were indeed false.
Block goes on to attribute to me “in conversation” the claim that training and reward did not influence the results. It is funny because we say it in the paper! But I did emphasize this at the pub after LeDoux and I gave a talk at the NYU philosophy of mind discussion group. Anyway, in response to that Block says that it would nullify the findings of the original paper that this is an effect of judgement. But that is silly because our claim was that since there is reason to think there is a difference in phenomenology and that the relevant difference psychologically/neurologically was a difference in HO representation then there is reason to think that HO state explains the difference in phenomenology.
Overall, then, I think it is really unfair to say that this argument is question begging. It does depend on their being an actual phenomenal difference when task performance is the same but we think we have good reasons to believe that which are independent of the higher-order view.
Richard Brown 
“It is too bad that Lau et al do not collect anecdotes from participants ….”
i suspect for this kind of psychophysics task when they do so many trials at near threshold, there isn’t a whole lot they can say anecdotally that wouldn’t be captured by the behavior already. this is not like a visual illusion type of demo where we can all marvel at it together, and discuss, compare notes, sort of thing.
but you’re right the everyday experience in the periphery should be sufficient, as far as anectodal evidence is concerned
I guess I was hoping there would some remarks made offhand by participants, like the ones we hear about from participants in the Sperling partial reports set up (the ones that seemingly hold a lot of water for those who like phenomenological overflow)…I am still hoping that one day someone will run these kinds of experiments and collect qualitative data (along with quantitative)…I tried to get some xphi people to do it a while back and nothing ever came of it…from my experience using these kinds of stimuli in the classroom over the years I would say that people really disagree about what they experience (in the Sperling case especially). In the full report case some say they felt like they saw all the letters, some say they felt like they only saw some of the letters, and ‘some other stuff’, while others claim to only have seen the letters they were focused don…in the partial report cases some people say they felt like they heard the cue and then quickly directed their attention to an image that was still there, while other deny that this is how they did the task…of course that is after just a few sub-optimally presented trials in a classroom…would be really interesting to see what most people say after getting good at it…in the case of your experiments I also think there would be something interesting to find out…say, after they become a bit practiced, give them a few trials (a small block) and then afterwards let them know which ones were false alarms and see if they were surprised.
Conveniently, Richard has given an argument in his response that nicely illustrates the question-begging form of argument that I was pinpointing. Here it is:
“It is too bad that Lau et al do not collect anecdotes from participants but I think just from our ordinary everyday experience we have some cases of inattention inflation. Sometimes as I am sitting at my computer writing something I will think that I saw the little red icon in the right corner of the screen that alerts me to an email in my inbox. Sometimes I will check and it will indeed be there. Other times I check and there is no red marker. But it sure did seem like there was one there just before I looked! The idea is that something like this is going on in the experimental conditions. I predict that if asked subjects would be surprised to find out that (some of) their false alarms were indeed false.”
As Richard and Hakwan note in their article, peripheral perception is more variable than focal perception (and the same kind of extra variability applies to inattentive perception and perception under trans-cranial magnetic stimulation). From a first order point of view, the explanation of the phenomenon that Richard points out could simply be due to a a strong perceptual representation which involves amplification of visual noise due to the higher variability. This is a straightforward first order explanation. An alternative first order picture of this case is that because of the variability of perception in the periphery, a peripheral signal is more likely to cross the confidence threshold for confidence that one saw it than a signal of equal strength in the fovea. So, the high level of confidence that Richard points to could then be explained as due to a difference in post-perceptual judgment rather than as due to a difference in consciousness. It is this difference in judgment that Richard appeals to in bolstering the higher order view. But we can see that his claim is question-begging since the phenomenon can be explained on the basis of either the first or higher order view.
Of course, on the first order view the best explanation of the difference in judgment is that the increased variability of first order representations in the periphery causes more high strength first order representations and those in turn are responsible for the differences in judgment. But it would be question-begging if I were to claim on this basis that the first order view wins. The right response is that “inattentional inflation” does not count one way or the other since it can equally well be explained as first order phenomenal inflation or as mere post-perceptual judgment inflation. The charge of question-begging against Richard is the same as the charge against the paper by Richard and Hakwan. They assert that the higher order view can explain phenomena involving the periphery instead of comparing their account with the kind of accounts first order theorists give.
Richard says this:
“In giving our answer to this Block edited a quote from us without indicating that in the text. We say “if a combined increase in the frequency of saying “yes I see the target” and higher visibility ratings is not good evidence that phenomenology has changed, what else can count?” and he quotes us as just saying if “higher visibility rating is not good evidence…” totally ignoring that we explicitly said it is the combination of both that we are replying on. This is misleading!
It is both of these that lead us to think that there really is a difference between the two cases and that leads us to think (B) is the right interpretation. They say they see it more often and also rate it as more visible even though they are not doing a better job of detecting the stimulus. It has nothing to do with the fact that we are willing to defend a higher-order approach to consciousness.”
In quoting, one always has to make a choice between completeness and streamlining. My reasoning here was first that Hakwan’s and Richard’s paper appears in the same volume and no one would be reading my reply without reading theirs. Second, in this context, there is no important difference between increased reports of seeing the target and increased visibility ratings. These are both expressions of judgments. In any case, the operative part of Richard’s reply is this: “They say they see it more often and also rate it as more visible even though they are not doing a better job of detecting the stimulus. It has nothing to do with the fact that we are willing to defend a higher-order approach to consciousness.”
Again, Richard commits the same fallacy of begging the question that I criticized. The increased likelihood of reporting seeing it and increased visibility ratings are indicative of increased strength of judgments. Higher order theory dictates increased consciousness but from the point of first order accounts, these are increases in post-perceptual strength that may or may not involve increased first order perceptual strength. It would be question-begging for either side to claim vindication from these data alone.
Hi Ned, thanks for responding, it is great to get your feedback on this stuff!
First, I think it is a mistake to say that Hakwan and I think the higher-order view is vindicated by these data alone! We are not committed to anything so strong as that. What I think (Hakwan can say if he disagrees with me) is that these data provide some evidence for the higher-order view. Roughly the line of thinking is that the interpretation of the results given by the higher-order view should be preferred over the interpretation given by the first-order view. We agree that there is a way to interpret the data from the first-order perspective but we just think that you have to do too much special pleading to make it work and so we should prefer the higher-order interpretation.
Speaking for myself I can say that I am not 100% convinced that the higher-order theory is true. I think of it is an under-appreciated option that I am willing to defend but for the record there are versions of a biological-first-order view that I would also defend. My over-arching commitment is to an identity theory of some kind and I have intuitions that tend to be split between higher-order and first-order approaches to physicalism…generally my views about the higher-order approach are that it is Not Obviously False and has Some Empirical Support and so is a Theory to Take Seriously. And that is compatible with there being other theories worth taking seriously.
So while I would be happy to be wrong about this, I don’t think I actually am 🙂
Ned says,
I agree that the anecdotal case I gave is open to this kind of interpretation, but part of our argument is that one of these options is ruled out by the data as an explanation for Inattention Inflation. We have some strong reasons to think that the first-order representations are the same in both cases. D’ is matched (and fMRI activity does not show an increase in activation in other studies, etc). If the difference in judgement were because of an actual difference in phenomenology (as Hakwan and I suggest) then we would expect to see that difference reflected in some kind of difference in activity in areas reasonably thought of as being first-order areas. But they did not find that. So the, by your own admission, best explanation of the data is at odds with what they found.
The other option (for FO views) is that it is merely post-perceptual inflation and this is the one that we say is outlandish. If the option is between that and the higher-order view then the higher-order view is the one that should be preferred. Yes, it is theoretically possible that this is the result of just an effect on post-perceptual judgement (even though it resists feedback, training, etc) and so on any given trial, there is no difference in what it is like between attended and unattended stimuli while subjects incorrectly judge that there was…but what reason (independently of FO theory) do we have to prefer that? It is not like in the TMS cases where you can say ‘oh but they knocked out the part of the brain that allows report’…here there seems to be no reason we would mistrust the participant’s reports and some reason to accept them so on balance it seems like we need some reason to think it goes your way instead of ours.
[…] On the other side of the spectrum, my paper with Hakwan Lau for the Ned Block volume was written a long time ago but just came out along with Block’s response (see my response here) […]