Single-case neuropsychological studies, such as those of H.M. and K.C., have been foundational in advancing our understanding of memory and cognition. However, the reliance on individual patients can introduce biases, as investigators may inadvertently emphasize findings that align with their hypotheses while neglecting alternative interpretations. For example, the focus on H.M.’s preserved remote memories led to decades of emphasis on consolidation models of memory, often at the expense of exploring alternative frameworks. This focus also delayed recognition of the hippocampus’s role in navigation and spatial cognition, which is now widely acknowledged.

Single-case studies are often fraught with challenges related to replicability and accessibility. By their nature, these studies cannot be replicated independently, making it difficult to confirm findings or test competing interpretations. Dominance by individual research groups, who may act as gatekeepers to a subject, can restrict access to data and hinder broader scientific collaboration. This approach runs counter to modern principles of open science, where data sharing and transparency are prioritized to foster collective scrutiny and progress. Without independent validation, conclusions drawn from single cases risk becoming entrenched, disproportionately shaping the direction of research.

While single-case studies provide invaluable insights, their findings should be interpreted with caution and situated within a broader context of evidence. Collaborative investigations, open data sharing, and attention to diverse hypotheses can reduce biases and enrich the interpretation of results. By complementing single-case findings with systematic approaches, such as group studies and meta-analyses, researchers can ensure a more comprehensive and balanced understanding of brain function, avoiding the pitfalls of overgeneralization and theoretical entrenchment.

There was a very interesting paper in the journal Nature this week concerned with an issue we recently discussed in class. The paper was by Sol and colleagues and concerned absolute and relative brain size and intelligence in a sample of 111 bird species. In this study, intelligence was defined by the propensity for innovation, a measure not that different from Deaner’s measure of novel problem solving that we discussed in Lecture 01 with regard to primate intelligence. Recall that in that study, Deaner found the intelligence in primates scaled with absolute brain size rather than relative brain size (i.e., brain size scaled by body size).

The present study shows that both absolute and relative brain size predicts the propensity for innovation in birds. Moreover, it was the number of neurons in the pallial regions (prosencephalon) that were most associated with this measure of intelligence. In addition, birds with larger brains when scaled by body size tended to have more pallial neurons. Thus, both measures contributed independently to the authors measure of bird intelligence.

In preparing my lectures for this course each year, I review the current literature and remove or qualify the discussion of topics where the foundational studies for those topics were challenged by newly published data, or where the original studies failed to replicate. That the scientific literature contains errors, and that attempts to replicate the findings of published studies frequently fail, is surprising to many who are new to science. However, failures to replicate and challenges to the published literature are common in all branches of science. Indeed, it is rare that any single experiment is definitive and scientists often live with uncertainty about key findings until sufficient numbers of publications by independent groups have confirmed the original results.

In the ideal, scientific questions are posed as a contest between two outcomes, or a test between two hypotheses. Let’s use a simple example typical of biomedical research: testing an inactive placebo and an active drug in lowering blood pressure. One hypothesis is the null, which specifies there is no difference in blood pressure when treated with either the placebo than the drug. The alternate hypothesis is that the active drug is more effective than the placebo in lowering blood pressure. The logic of the experiment is designed to evaluate the null hypothesis; i.e., that there is no difference between the active and inactive compounds. One rejects that null hypothesis when the evidence is statistically overwhelming that the active drug does a better job of lowering blood pressure than the placebo. This seems to be straight-forward, so how do incorrect results get into the published literature?

There are many possibilities. One possibility is that blood pressure varies day-to-day for intrinsic reasons unrelated to the experiment, and this variation, or ‘noise’, is a major contributor to the measurement of blood pressure. Perhaps the real effect of the active drug was smaller than this day-to-day variability, and so the null hypothesis was accepted when it should have been rejected. Or, it could have been that too few subjects were tested and, a few subjects had randomly low readings unrelated to the active drug when blood pressure was measured after active drug treatment. This may have caused the experimenter to erroneously reject the null hypothesis and conclude that the active drug was working. Small effect sizes (how much the true effect is larger than intrinsic noise) and small and inadequate sample sizes are major contributors to incorrect results being published.

Another set of reasons for publication of incorrect results is that the experiment used poor experimental design or included confounding variables. Perhaps blood pressure was measured in the morning following active drug treatment and in the evening for the placebo treatment, and thus introduced a systematic time-of-day confound. Perhaps all of the subjects in the active drug treatment conditions were younger than the subjects in the active drug treatment. You get the idea.

Finally, it is painfully true that some published studies are biased towards particular outcomes. It has been found that experiments supported financially by pharmaceutical companies more often report positive effects of experimental drugs than do independently funded studies. This may reflect the selective publication of positive drug effects (positivity bias) and/or the absence of publication of negative results (the ‘file-drawer’ problem). Unfortunately, it can also result from experimental malfeasance.

Regardless of the reasons, science (again, in the ideal) should be self-correcting. That is, studies by independent groups that test the same drugs may, or may not, replicate the initial studies. Over time, errors become corrected by the accumulating weight of the evidence, and scientific progress continues. Or does it?

There are severe impediments to the self-correcting nature of science. One is the sheer volume of scientific papers. A NSF worldwide survey of all science publications revealed more than 2.5M articles were published in all disciplines in 2018 alone. Another impediment is that scientific journals prioritize publishing positive results over negative results, reasoning that scientists don’t want to read about experiments that failed. This results in the aforementioned file-drawer problem where negative results (i.e., those that do not reject the null hypothesis) are never published but rather sit in the investigator’s file drawer. Finally, there is the (accurate, I believe) perception among scientists that publishing negative results does not lead to career advancement.

My own opinion is that science remains self-correcting, but that the pace of correction is slow and, perhaps, slowing. And once a positive result is published, it takes great effort to dislodge it. A single non-replication is often not enough.

One recent example of correction in science concerns the efficacy of hydroxychloroquine (an anti-malaria drug) in the prevention and treatment of Covid–19. The specific issue of hydroxychloroquine, and the general issue of self-correction in science, was reported at the online health news website Statnews. The following summary is taken from their reporting.

The original report of positive results of hydroxychloroquine for treating Covid was published in the International Journal of Antimicrobial Agents after first appearing in mid-March 2020 on a non-peer reviewed preprint server. The peer review process was concluded in a single day (most journals take months for peer review). The study itself was a small sample (20 patients with Covid–19 who received hydroxychloroquine and 16 controls), non-randomized, open-label study. According to Statnews, independent researchers had raised extremely serious concerns with the study’s methods and conclusions with days after its publication. A subsequent independent review concluded that “this study suffers from major methodological shortcomings which make it nearly if not completely uninformative.” Despite this, the paper excited great interest and millions of individuals subsequently received hydroxychloroquine as a treatment for Covid–19.

Subsequent large-scale and well-controlled studies of hydroxychloroquine confirmed that it was useless as a preventative measure or treatment for Covid–19. Indeed, meta-analysis reported in Nature Communications that considered data from more than 10,000 patients demonstrated that patients with Covid–19 receiving hydroxychloroquine as a treatment were more likely to die than patients with Covid–19 who were not given this drug. Why this is so is not clear, but may reflect bias in who received hydroxychloroquine treatment (for example, perhaps the drug was administered only to the most severe cases). So, ultimately, science self-corrected but at great cost in time, money, and, perhaps, lives.

A telling coda to this story is that the journal article that first reported the value of hydroxychloroquine was cited 4000 times. The review paper that ‘corrected’ the science was only cited 38 times (both figures from Google Scholar, as reported by Statnews).

The case of hydroxychloroquine, though illustrative, is extreme. Most issues of scientific self-correction do not involve life or death issues. However, there are many instances where unsubstantiated scientific reports affect public policy. We will discuss some of those instances in the realm of neuroscience throughout the semester.

The physicist and Nobel laureate Max Planck offered the following observation about scientific progress:

A new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it.

A pithy restatement of Planck’s observation is that science advances one funeral at a time. But is this really true? Azoulay and colleagues attempted to find out and presented their results in a 2019 study in the American Economic Review. These authors developed a clever approach by which they investigated the consequences of the premature death of a scientific superstar on journal publication rates within the subfield in which that scientist worked. The study was restricted to the life sciences where publications by the deceased scientist, that scientist’s collaborators and co-authors, and by other unrelated scientists could be easily quantified by subfield.

The authors created a reasonable set of criteria to identify ‘scientific superstars’ that could be consistently applied. For example, the criteria included a group of prestigious awards and grants and included elections to memberships in organizations such as the National Academy of Science. Using these criteria, a total of 452 eminent scientists were identified for study.

The authors concluded that the premature death of a scientific superstar opens up that scientist’s field to innovation. Following the death of the superstar scientist, there was a modest increase in the number of publications in the deceased scientist’s subfield. However, this overall increase represented two opposing trends. The first trend was that the number of publications in the subfield by the deceased scientist’s collaborators and co-authors went down dramatically. The second trend was that publications in the subfield by scientists who were not collaborators of the deceased scientist went up. The large increase in publications within the subfield by unrelated scientist more than made of the difference.

The authors further noted that the increase in subfield publications did not represent a reshuffling of the productivity of scientists who were already in the subfield. Rather, it represented and influx from scientists from other subfields moving into the subfield of the deceased scientist.

The authors suggest that the scientific superstar and collaborators form a clique that imposes a high cost of entry into the subfield in which the superstar dominates. This cost of entry might be reflected by allocation of grant funding by study sections dominated by the superstar’s collaborators (so-called financial gatekeeping), by greater ease of publication in scientific journals for clique members, and by invitations to speak at conferences. Scientists who are not members of this clique might decide that the cost of entry into that field is too high, until the death of the superstar scientist creates an opening.

The effects observed by Azoulay and colleagues were greatest in subfields that were losing scientific momentum, and that the contributions of the scientists newly attracted to the subfield rejuvenate the field by moving its ‘intellectual center of gravity’ away from its position prior to the death of the scientific superstar, particularly by introducing new ideas and methods from other fields.

While the study presents the pessimistic view that scientific superstars and their clique of collaborators retard scientific progress by restricting entry of new ideas and new scientists into their field, the authors point out that this dominance may have been important in the field’s inception. That is, for a new subfield to flourish initially, it may have been important for the founding scientist to control its intellectual evolution so that progress could be made through ‘shared assumptions and methodologies’. However, over time, this dominance chokes further progress.

Azoulay, P., Fons-Rosen, C., & Zivin, J. S. G. (2019). Does Science Advance One Funeral at a Time. Am Econ Rev, 109(8), 2889-2920.

Several theories have considered language to have evolved from tool use. In a recent article in the journal Science, Thibault and his colleagues tested their theory that tool use requires ‘…integrating an external object as a body part and embedding its functional structure in the motor program’. In their view, this adds a ‘hierarchical level into the motor plan’ that involves ‘modifying relationships among interdependent subcomponents’. Thibault and colleagues noted that ‘embedded structures also exist in language’ and that syntax is the function used to organize linguistic hierarchies. Thus, they argue that syntax is a function common to both language and tool use and conducted two experiments to test this hypothesis.

In the first experiment, subjects performed both a syntax task and a novel tool use task while brain images were acquired using functional magnetic resonance imaging (fMRI). This imaging modality captures blood oxygenation changes in the brain and serves as a proxy measure for neuronal activation. In the tool use task, subjects had to use a long tong-like tool to precisely move small items on a board. In the syntax task, subjects had to read sentences with complex syntax (e.g., The writer that the poet admires writes the paper) and then select a true statement about the sentence from among four possibilities (e.g., The writer writes the paper). These tasks were first compared simple syntax and simple movement tasks, and then to each other. The main finding was that a small region of the basal ganglia was commonly activated by both tool use and language syntax.

It is notable that the overlapping activation occurred in the basal ganglia and not in the left inferior frontal gyrus (IFG). The left IFG is the location of Broca’s area which when damaged in humans causes an expressive aphasia and difficulty in processing syntax. The left IFG is also considered by some investigators to be the location in humans that is analogous the mirror neuron area in non-human primates. Mirror neurons are those that respond both to motor actions performed by the agent, but also in response to observing those actions performed by another agent. Thibault and colleagues observed activation in the left IFG for both the syntax and the tool use task, but the activations were spatially distinct and did not overlap.

In the second experiment, Thibault and colleagues determined whether the shared circuitry between the two tasks was amenable to cross training. Two groups of subjects were tested. One group was trained on the language syntax task and then tested with the novel tool use task. The other group was trained on the novel tool use task and tested on the syntax task. In both groups, training on one task improved performance on the other task. This supported their overall conclusion that a common neural circuitry was being used in both tasks.

Students interested in learning more about theories linking tool use and language can consult a 2012 special issue of the Philosophical Transactions of the Royal Society devoted to this interesting topic.

References:

Thibault, S., Py, R., Gervasi, A. M., Salemme, R., Koun, E., Lövden, M. et al. (2021). Tool use and language share syntactic processes and neural patterns in the basal ganglia. Science, 374(6569).

Steele, J., Ferrari, P. F., & Fogassi, L. (2012). From action to language: comparative perspectives on primate tool use, gesture and the evolution of human language. Philos Trans R Soc Lond B Biol Sci, 367(1585), 4-9.

A recent paper in Nature Immunology provides additional evidence for the role of cytokine signaling molecules and behavior.

An accessible summary of the article was published in The Scientist magazine. The study considered immune system T-cells and the cytokine interleukin in the meninges of the brain. The investigators found that mice with these T-cells and cytokines in their meninges showed more cautious behavior in experimental open-arm mazes than those mice without these immune cells and signaling molecules. In addition, the scientists demonstrated a relationship between the activity in these immune and signaling molecules and gut bacteria. Studies of human cadavers have also revealed the presence of these molecules in the meninges of the human brain, raising the possibility that these mechanisms may also exist in humans. This is one of a number of recent papers demonstrating the close link between our immune and nervous systems.

In a NY Times article published on September 9th, Yale immunologist Dr. Akiko Iwasaki explained what happens when the SARS-CoV-2 virus attacks the brain. Her work showed that the coronavirus ‘…exploits the brain cells’ machinery to multiply, but doesn’t destroy them. Instead, it chokes off oxygen to adjacent cells, causing them to wither and die’. The virus also appears to decrease the number of synapses in the brain. Included with the article, but from a different study, are a series of MRI scans of individuals who had SARS-CoV-2 brain infections. These scans show areas of infarction.

Brain-Computer Interfaces (BCI)

A recent article in the NY Times highlights the research of Dr. Jack Gallant at UC Berkeley, who was among the first who applied computational approaches to decoding the visual images that a subject was viewing from functional MRI data. Gallant received his Ph.D. in 1986 from Yale’s Psychology department. This research provides hope to many individuals who have suffered disease or trauma and have lost their ability to move their limbs or speak. In each of these realms, BCIs have shown promise as intentions expressed in neural activity can be read by computers which then control complex robotic arms and speech synthesizers. However, this work also raises serious ethical concerns if reading neural activity provides others with information an individual does not want to share. Many tech companies such as Google and Facebook and Elon Musk’s NeuraLink are investing in such technology, which has prompted an urgent examination of the ethics of mind reading.

Recent research has focused attention on microglia, a type of glial cell that is associated with the immune system, and which is normally activated by infection or damage to the brain. Recent research has shown that microglia also act as synaptic strippers and in a 2020 paper in Science, Wang and colleagues showed that microglia can remove connections between hippocampal neurons that are responsible for particular memories. This suggests that microglia play an important role in forgetting.

An interesting article recently appeared in Nature Communications concerning voluntary exercise in mice. The authors deleted a gene encoding a particular protein the the AgRP neurons of the hypothalamus. This deletion caused a the animal to reduce voluntary exercise and become obese.