I’m very pleased to announce our latest Mind & Language symposium on Matthias Michel and Jorge Morales’ forthcoming “Minority Reports: Consciousness and the Prefrontal Cortex.” Our outstanding commentators on the target article include Liz Irvine (Cardiff), Benjamin Kozuch (Alabama), and Michael Pitts with Kevin Ortego (Reed College).
***
A great deal of neuroscience research focuses on finding neural correlates of consciousness (NCCs), that is, neural states that are minimally sufficient for mental states to be conscious (Chalmers 2000). A number of neuroscientists and philosophers have maintained on the basis of experimental and theoretical evidence that prefrontal cortex is a NCC. Examples include the global workspace theory (Baars 2005, Dehaene & Changeux 2011) and the higher-order theory (Brown 2015, Lau & Rosenthal 2011, Lau & Brown 2019). Proponents of “back of the head” theories, in contrast, argue that consciousness primarily depends on neural activity in posterior parts of the cortex (Lamme 2006) or a “posterior hot-zone” (Koch et al. 2016, Koch 2018).
Numerous experiments, reviewed by Michel and Morales, provide evidence that conscious perception elicits increased activity in the PFC, while unconscious perception does not. Proponents of prefrontal theories take this to be evidence that PFC is part of the neural substrate of consciousness. Most of the experiments to which proponents of prefrontal theories appeal, however, make use of a “report-based” experimental method in which subjects are directed to report on whether they consciously perceive a stimulus or not. According to what Michel and Morales call the “argument from report,” this introduces a confound: the presence of neural activity in PFC that is correlated with conscious perception is also correlated with post-perceptual cognitive processes that enable verbal report. Hence, methods that require subjects to provide explicit reports of awareness (and unawareness) should be eliminated from analyses that search for the NCCs.
Michel and Morales defend prefrontal theories against the argument from report. In section 2 of the target article, they argue that in experiments comparing unconscious and conscious conditions, the observed difference in brain activity in the PFC is unlikely to be explained solely by its role in generating reports. “[T]here is little reason to believe,” they suggest, “that ‘seen’ and ‘unseen’ reports involve fundamentally different cognitive capacities.”
Proponents of the argument from report predict that in no-report experimental paradigms, differences in PFC activity between conscious and unconscious trials should be negligible. In section 3, Michel and Morales argue that this prediction conflicts with the best interpretation of the available evidence: there are differences in PFC activity between conscious and unconscious trials, they suggest, even when subjects do not have to provide reports.
***
Comments on this symposium will be open for at least a couple of weeks. Many thanks to Matthias, Jorge, and our three commentators for their stellar contributions! All of us here at the Brains Blog are also grateful to Gregory Currie, the other Mind & Language editors, and the staff at Wiley-Blackwell for their continued support of these symposia.
- You can learn more about Matthias Michel and his research here.
- You can learn more about Jorge Morales and his research here.
Below there are links to the authors’ video introduction, the target article, commentaries, and the authors’ replies.
1. Video Introduction by Matthias Michel and Jorge Morales:
2. Target Article: “Minority reports: Consciousness and the Prefrontal Cortex”
3. Commentaries and Replies:
- Liz Irvine,“Commentary on Minority Reports”
- Benjamin Kozuch, “Minority Retorts: Other Explanations for the Prefrontal Correlations”
- Michael Pitts with Kevin Ortego, “Why ‘No-Report’ Paradigms are an Important Tool for Consciousness Research”
- Matthias Michel and Jorge Morales, Replies
Thanks, Matthias and Jorge, for giving my comments some thought, and taking the time to address some of the objections I had to the idea that the PFC is an NCC. Even if that wasn’t the immediate concern of your paper, I would guess that some people might want to take your rather well-constructed arguments to add to the overall support for PFC consciousness (perhaps even you!?), and so it is interesting ask how well your arguments might do this…
There’s probably a few different things I might say about your response to my arguments against PFC consciousness, but I’ll mention just one. It concerns my argument that we shouldn’t expect the PFC to be the content NCC for sensory consciousness, the reason here being because this would involve the PFC having to represent all of the rich, varied, and numerous sensory properties that we experience at any given moment, something that it would be hard to find evolutionary justification for. You take it to be significant that sensory content can be decoded from PFC activity. But I am not sure how well there being *some* sensory content in the PFC (even if it matches the content found in consciousness) helps to undermine the concern here, which is that there is a lack of evolutionary justification for there being a *very large* amount of sensory content (i.e., the amount we find in the average, multi-modal sensory experience) being re-represented in executive parts of the cognitive architecture (i.e., in this case, the PFC). In any case, it’s not as if we should be surprised to find a bit of sensory content in the PFC, given that it is known to be involved in the metacognition of sensory states. So I would submit that the ability to locate *some* sensory content in the PFC, in and of itself, might not do a very good job of blunting the evolutionary justification worry.
Hi Benji,
Thanks for your comment! I agree with you that if consciousness is rich, then it’s difficult for re-representationalist PFC theories to explain why all this sensory content should be duplicated in PFC. This, however, requires that you accept that consciousness is rich in the first place. PFC-re-representationalists (we should definitely find a shorter label) would probably deny that. As you know there are many mechanisms that PFC theorists can appeal to in order to explain the apparent richness of conscious experiences while also saying that PFC actually re-represents only a limited amount (e.g. ensemble representations, inflation, filling-in, and all that good stuff). And I don’t think it’s a problem to explain the evolutionary purpose of re-representations if consciousness is sparse (see e.g. Joe LeDoux’s recent book on the evolution of consciousness, in which he defends a re-representationalist view).
Even assuming that they accept that consciousness is rich, I’m pretty sure it’s possible to find a just-so story here. And it’s difficult to have more than just-so stories since, first, we don’t have really good theories about the evolutionary functions of consciousness, and, second, determining which non-human animals are conscious (or not) is a mess. In a field where evolutionary reasoning might be so unconstrained (since we don’t know much at all), I’m not really sure about the value of reverse engineering arguments. I like those arguments but we have to admit that it’s *really* speculative at the moment. I nevertheless agree with you that some stories might sound prima facie more or less plausible and that, all other things being equal, re-representationalist theories would sound less plausible *if* consciousness is rich.
Thanks again for your comment and for participating in this!
Matthias
Matthias,
Thanks for your insightful response, which I think really helped to sharpen the lines of the debate over PFC consciousness. Just thought I’d mention, in way of a brief reply, that the idea that consciousness is not rich might be further from being established than you think. For instance, some of the data used to support “subjective inflation” seems to me to fail to establish that there is less content in visual experience than we thought, but rather to establish only that we are sometimes mistaken about stimuli. As way of example, consider how the Solevey et al study (2015) only shows subjects to be mistaken about whether there is a grating or noise in peripheral vision, not that they lacked any experience at all in that part of the visual field. (We could guess they wouldn’t report this to be the case.) Similarly, if a part of the visual field is “filled in,” one is still experiencing something there (i.e., there is again no lack of experience in that part of the visual field). The point of this would be that these sorts of data and ideas don’t seem get rid of the need for the PFC to be hypothesized as representing a lot of sensory content, but rather just show the PFC perhaps to be representing *different* content.
In any case, I feel like there has been some progress made here, insofar as it seems that we have come to an agreement that the health of duplicative theories of PFC consciousness (i.e., “re-representationalist”) might very well depend on the ability to make plausible the idea of us having a “thin” sensory experience.
Well, thanks again for the opportunity to participate in this symposium on your high-quality piece of empirical philosophy.
-Benji
General Comments
We very much appreciate Michel & Morales’s (M&M’s) reply to the commentaries. In their reply, M&M clearly articulate the three mains points they aimed for in their original paper. This was necessary because, according to M&M, two out of three of the commentaries (including ours) misinterpreted their main messages. In their original paper, M&M appeared to argue that report paradigms are already well-controlled (therefore, no-report paradigms are unnecessary) and results from both report and no-report paradigms clearly indicate that PFC plays a direct role in conscious perception. We disagreed with both of these points and felt that such arguments sent the wrong message, especially to young scientists in the field, which is largely why we agreed to write a commentary in the first place. Thankfully, M&M explain that this was a misreading of their original message, and in their reply to commentaries, they summarize the three main points they intended to make. Here we will consider these three points one-by-one.
First, we completely agree with their point about the “simple version of the argument from report”. So-called “no-report” paradigms should not only focus on eliminating reports (buttons presses or verbal responses), but should aim to reduce or eliminate all of the extra cognitive processing that ensues after a stimulus is consciously perceived. This has been obvious all along to those of us working on developing no-report paradigms, but it may not have been obvious to others in the field, so we appreciate that M&M articulated this so clearly in writing.
Their second point is about the importance of identifying particular cognitive processes that may be confounded with conscious perception in report-based paradigms. Here, we only partially agree. In a separate comment (titled “Isolating confounds vs. isolating NCCs”), we will outline our disagreements.
Regarding their third point, we completely agree that no-report paradigms so far have not conclusively shown that PFC theories of consciousness are wrong, and that some of the claims made in recent review papers may have been premature and overstated. However, we continue to disagree with M&M’s argument about the interpretation of null results and provide an additional comment on this point as well (titled “Null results”).
Overall, we appreciate the improved clarity in M&M’s reply to commentaries and we hope that interested readers will focus on their reply in order to avoid misinterpreting their original paper.
-Michael Pitts & Kevin Ortego
Isolating confounds vs. isolating NCCs
In M&M’s reply to commentaries, their second main point is that “proponents of no-report paradigms should be clear on what is supposed to constitute a confound and why”. While we agree that it is interesting and useful for the field of cognitive neuroscience to dissect every component of cognitive processing that occurs when humans perform a task, such a strategy seems unnecessary and inefficient when the goal is to isolate the neural mechanisms of conscious perception.
A simpler, more efficient approach is to eliminate as many post-perceptual cognitive processes as possible in a no-report condition in order to rule-out the associated brain activity as being critical for conscious perception. If we can clearly show that a given brain signal completely disappears in a no-report condition, despite us making the same contrast between consciously seen vs. unseen stimuli, it seems as though we have made progress in isolating NCCs, or at least in ruling out signals that are unlikely to be NCCs. Such progress is possible without having to spend years teasing apart which exact cognitive processes were confounding our measurements in report conditions.
In their reply, M&M argue that proponents of no-report paradigms must first determine whether the extra cognitive processing that occurs in report conditions is associated with working memory encoding vs. working memory maintenance vs. object categorization vs. judgment and decision-making vs. sub-vocal labeling vs. associative thinking vs. motor planning vs. anything else. In our view, the burden of understanding and controlling for every particular cognitive process that might follow conscious perception should be on proponents of report paradigms, since some of these processes will likely *always* be present in *any* paradigm that requires trial-by-trial reports. This is one of the reasons we are so excited about adding no-report paradigms to our methodological toolkit. If we carefully design no-report paradigms (ensuring that the seen vs. unseen contrast remains valid), we can simply rule-out all extra brain activity patterns that were unique to the report condition without having to understand which exact cognitive processes were associated with these activity patterns.
To be clear, we are not opposed to the approach that M&M suggest, since improving our understanding of human cognition is interesting and important in its own right. However, for the particular goals of the NCC project, we think it’s much more efficient and useful to focus on incorporating no-report paradigms that can clearly rule-out extra cognitive processes regardless of the particular types of processes we are ruling out. We should also keep in mind that certain cognitive confounds may still be present in no-report conditions, and it will be important to develop a variety of no-report protocols to “triangulate” the NCC (as M&M aptly suggest in their reply).
Null results
How should we interpret “null results” or “negative findings”, both in general, and in the context of PFC activity or the P3b? In our view, interpreting null results is equally as complicated as interpreting “statistically significant results” or “positive findings”. Our main argument here is that what’s most important is to critically evaluate all of the details of individual experiments, including potential confounds, control conditions, the magnitude of effects, their replicability, and the over-all pattern of results which often includes a combination of positive and negative findings.
For example, in all of our previous experiments that include no-report conditions, we find that a very large brain signal that is present and robust in report conditions, the P3b (often 8-16uV in magnitude), completely disappears in no-report conditions. At the same time, a much smaller event-related potential, the so-called “visual awareness negativity” or “VAN” (typically 1-2uV in magnitude), remains stable in both report and no-report conditions. It seems incorrect to interpret this overall pattern of results as a “null finding”. M&M’s analogy about not detecting Saturn because our telescope is too small just doesn’t work in this situation. Why is our telescope perfectly adequate for detecting brain signals 1/8 of the size of the P3b, but is inadequate for detecting the P3b in no-report conditions? It remains to be seen whether similar arguments can be made for fMRI studies employing no-report paradigms. Do robust PFC activations in report conditions disappear under no-report conditions, while weaker activations in other brain regions remain stable regardless of the reporting task? Such a pattern may be helpful in assessing the sensitivity of the brain measurements and the potential meaningfulness of “null” findings in PFC.
More generally, when formulating a working hypothesis (or a current belief based on incomplete evidence), we should take into consideration all of the available evidence, including both positive and negative findings, and update our beliefs according to the entire pattern of evidence. Instead of pre-judging the informativeness of a result based on whether it is a positive or negative finding, we should look at the details of each study, including the factors that were controlled for, the potential flaws in each part of the design, whether alternative explanations exist and how likely each alternative might be, what the overall pattern of results suggests both within and across studies, etc. A solid set of null findings can be just as informative as a solid set of positive findings if we are asking the right questions, designing our experiments carefully, and including sensitivity-checks in our analyses.