An Amicable Discussion About Psychology and Neuroscience

People like conflict (the interpersonal kind, not BLUE).¹ Or at least, they like scientific debate at conferences. Panel discussions that are too harmonious seem to be divisive. Some people will say, “well, now THAT wasn't very controversial.” But as I mentioned last time, one highlight of the 2019 Cognitive Neuroscience Society Annual Meeting was a Symposium organized by Dr. David Poeppel.²

Special Session - The Relation Between Psychology and Neuroscience, David Poeppel, Organizer, Grand Ballroom

Whether we study single cells, measure populations of neurons, characterize anatomical structure, or quantify BOLD, whether we collect reaction times or construct computational models, it is a presupposition of our field that we strive to bridge the neurosciences and the psychological/cognitive sciences. Our tools provide us with ever-greater spatial resolution and ideal temporal resolution. But do we have the right conceptual resolution? This conversation focuses on how we are doing with this challenge, whether we have examples of successful linking hypotheses between psychological and neurobiological accounts, whether we are missing important ideas or tools, and where we might go or should go, if all goes well. The conversation, in other words, examines the very core of cognitive neuroscience.

Conversation. Not debate. So first, let me summarize the conversation. Then I'll get back to the merits (demerits) of debate. In brief, many of the BIG IDEAS motifs of 2017 were revisited...

• David Marr and the importance of work at all levels of analysis 
• computation – algorithm – implementation
• What are the “laws” that bridge these levels of analysis?
• “Emergent properties” – a unique higher-level entity (e.g., consciousness, a flock of birds) emerges from the activity of lower-level activity (e.g., patterns of neuronal firing, the flight of individual birds)... the sum is greater than its parts
• Generative Models – formal models that make computational predictions

...with interspersed meta-commentary on replication, publishing, and Advice to Young Neuroscientists. Without further ado:

Dr. David Poeppel – Introductory Remarks that examined the very core of cognitive neuroscience (i.e., “we have to face the music”).

• the conceptual basis of cognitive neuroscience shouldn't be correlation 

For example, fronto-parietal network connectivity (as determined by resting state fMRI) is associated with some cognitive function, but that doesn't mean it causes or explains the behavior (or internal thought). We all know this, and we all know that “we must want more!” But we haven't the vaguest idea of how to relate complex psychological constructs such as attention, volition, and emotion to ongoing biological processes involving calcium channels, dendrites, and glutamatergic synapses.

• but what if the psychological and the biological are categorically dissimilar??

In their 2003 book, Philosophical Foundations of Neuroscience, Bennett and Hacker warned that cognitive neuroscientists make the cardinal error of “...commit[ting] the mereological fallacy, the tendency to ascribe to the brain psychological concepts that only make sense when ascribed to whole animals.”

• cognitive neuroscientists (gasp) are actually Cartesian dualists! Bennett and Hacker:

“For the characteristic form of explanation in contemporary cognitive neuroscience consists in ascribing psychological attributes to the brain and its parts in order to explain the possession of psychological attributes and the exercise (and deficiencies in the exercise) of cognitive powers by human beings.” (p. 3)

On that optimistic note, the four panelists gave their introductory remarks.

(1) Dr. Lila Davachi asked, “what is the value of the work we do?” Uh, well, that's a difficult question. Are we improving society in some way? Adding to a collective body of knowledge that may (or may not) be the key to explaining behavior and curing disease? Although still difficult, Dr. Davachi posed an easier question, “what are your goals?” To describe behavior, predict behavior (correlation), explain behavior (causation), change behavior (manipulation)? But “what counts as an explanation?” I don't think anyone really answered that question. Instead she mentioned the recurring themes of levels of analysis (without invoking Marr by name), emergent properties (the flock of birds analogy), and bridging laws (that link levels of analysis). The correct level of analysis is/are the one(s) that advance your goals. But what to do about “level chauvinism” in contemporary neuroscience? This question was raised again and again.

(2) Dr. Jennifer Groh jumped right out of the gate with this motif. There are competing narratives in neuroscience we can call the electrode level (recording from neurons) vs. the neuroimaging level (recording large-scale brain activations or “network” interactions based on an indirect measure of neural activity). They make different assumptions about what is significant or worth studying. I found this interesting, since her lab is the only one that records from actual neurons. But there are ever more reductionist scientists who always throw stones at those above them. Neurobiologists (at the electrode level and below) are operating at ever more granular levels of detail, walking away from cognitive neuroscience entirely (who wants to be a dualist, anyway?). I knew exactly where she was going with this: the field is being driven by techniques, doing experiments merely because you can (cough — OPTOGENETICS — cough). Speaking for myself, however, the fact that neurobiologists can control mouse behavior by manipulating highly specific populations of cells raises the specter of insecurity... certain areas of research might not be considered “neuroscience” any more by a bulk of practitioners in the field (just attend the Society for Neuroscience annual meeting).

(3) Dr. Catherine Hartley continued with the recurring theme that we need both prediction and explanation to reach our ultimate goal of understanding behavior. Is a prediction system enough? No, we must know how the black box functions by studying “latent processes” such as representation and computation. But what if we're wrong about representations, I thought? The view of @PsychScientists immediately came to mind. Sorry to interrupt Dr. Hartley, but here's Golonka and Wilson in Ecological Representations:

Mainstream cognitive science and neuroscience both rely heavily on the notion of representation in order to explain the full range of our behavioral repertoire. The relevant feature of representation is its ability to designate (stand in for) spatially or temporally distant properties ... While representational theories are a potentially a powerful foundation for a good cognitive theory, problems such as grounding and system-detectable error remain unsolved. For these and other reasons, ecological explanations reject the need for representations and do not treat the nervous system as doing any mediating work. However, this has left us without a straight-forward vocabulary to engage with so-called 'representation-hungry' problems or the role of the nervous system in cognition.

They go on to invoke James J Gibson's ecological information functions. But I can already hear Dr. Poeppel's colleague @GregoryHickok and others on Twitter debating with @PsychScientists. Oh. Wait. Debate.

Returning to The Conversation that I so rudely interrupted, Dr. Hartley gave some excellent examples of theories that link psychology and neuroscience. The trichromatic theory of color vision — the finding that three independent channels convey color information — was based on psychophysics in the early-mid 1800s (Young–Helmholtz theory). This was over a century before the discovery of cones in the retina, which are sensitive to three different wavelengths. She also mentioned the more frequently used examples of Tolman's cognitive maps (which predated The Hippocampus as a Cognitive Map by 30 years) and error-driven reinforcement learning (Bush–Mosteller [23, 24] and Rescorla–Wagner, both of which predate knowledge of dopamine neurons). To generate good linking hypotheses in the present, we need to construct formal models that make quantitative predictions (generative models).

(4) Dr. Sharon Thompson-Schill gave a brief introduction with no slides, which is good because this post has gotten very long. For this reason, I won't cover the panel discussion and the Q&A period, which continued the same themes outlined above and expanded on “predictivism” (predictive chauvinism and data-driven neuroscience) and raised new points like the value (or not) of introspection in science. When the Cognitive Neuroscience Society updates their YouTube channel, I'll let you know. Another source is the excellent live tweeting of @VukovicNikola. But to wrap up, Dr. Thompson-Schill asked members of the audience whether they consider themselves psychologists or neuroscientists. Most identified as neuroscientists (which is a relative term, I think). Although more people will talk to you on a plane if you say you're a psychologist, “neuroscience is easy, psychology is hard,” a surprising take-home message.


Debating Debates

I've actually wanted to see more debating at the CNS meeting. For instance, the Society for the Neurobiology of Language (SNL) often features a lively debate at their conferences.³ Several examples are listed below.

2016:
Debate: The Consequences of Bilingualism for Cognitive and Neural Function
Ellen Bialystok & Manuel Carreiras

2014:
What counts as neurobiology of language – a debate
Steve Small, Angela Friederici

2013: Panel Discussions
The role of semantic information in reading aloud
Max Coltheart vs Mark Seidenberg

2012: Panel Discussions
What is the role of the insula in speech and language?
Nina F. Dronkers vs Julius Fridriksson


This one-on-one format has been very rare at CNS. Last year we saw a panel of four prominent neuroscientist address/debate...
Big Theory versus Big Data: What Will Solve the Big Problems in Cognitive Neuroscience?

.@gallantlab "This debate is silly" #CNS2018 pic.twitter.com/cpT7tKI0xa

— CNS News (@CogNeuroNews) March 24, 2018

Added-value entertainment was provided by Dr. Gary Marcus, which speaks to the issue of combative personalities dominating the scene.⁴

Gary Marcus talking over Jack Gallant. Eve Marder is out of the frame.

image by @CogNeuroNews

I'm old enough to remember the most volatile debate in CNS history, which was held (sadly) at the New York Marriott World Trade Center Hotel in 2001. Dr. Nancy Kanwisher and Dr. Isabel Gauthier debated whether face recognition (and activation of the fusiform face area) is a 'special' example of domain specificity (and perhaps an innate ability), or a manifestation of plasticity due to our exceptional expertise at recognizing faces:

A Face-Off on Brain Studies / How we recognize people and objects is a matter of debate
. . .

At the Cognitive Neuroscience Society meeting in Manhattan last week, a panel of scientists on both sides of the debate presented their arguments. On one side is Nancy Kanwisher of MIT, who first proposed that the fusiform gyrus was specifically designed to recognize faces–and faces alone–based on her findings using a magnetic resonance imaging device. Then, Isabel Gauthier, a neuroscientist at Vanderbilt, talked about her research, showing that the fusiform gyrus lights up when looking at many different kinds of objects people are skilled at recognizing.

Kudos to Newsday for keeping this article on their site after all these years.


Footnotes

¹ This is the color-word Stroop task: name the font color, rather than read the word. BLUE elicits conflict between the overlearned response ("read the word blue") and the task requirment (say "red").

² aka the the now-obligatory David Poeppel session on BIG STUFF. See these posts:

• The Big Ideas in Cognitive Neuroscience, Explained #CNS2017
• Big Theory, Big Data, and Big Worries in Cognitive Neuroscience #CNS2018

³ Let me now get on my soapbox to exhort the conference organizers to keep better online archives  — with stable urls — so I don't have to hunt through archive.org to find links to past meetings.

⁴ Although this is really tangential, I'm reminded of the Democratic Party presidential contenders in the US. Who deserves more coverage, Beto O'Rourke or Elizabeth Warren? Bernie Sanders or Kamala Harris?