The heritability of reinforcement learning parameters and their association with anxiety

In a large twin study, researchers found that while extinction learning rates robustly predict anxiety severity, neither extinction nor safety learning rates are heritable or genetically linked to anxiety, suggesting they are not endophenotypes despite the confirmed heritability of safety learning parameters.

The Gist

Imagine your brain is a highly sophisticated learning computer that constantly updates its "danger map" based on new experiences. When something scary happens, the computer marks it as a threat. But when that same thing turns out to be harmless, the computer needs to update its map to say, "Okay, this is actually safe now."

This study was like a massive family reunion investigation involving 925 twins. The researchers wanted to see if the way people's brains update these "danger maps" is something we inherit from our parents (like eye color) and if these specific learning styles are linked to anxiety.

Here is what they found, broken down into simple analogies:

1. The Two Types of Learning

The study looked at two specific ways the brain learns:

• Safety Learning: This is like learning that a dog you never met before is actually friendly. You meet it, realize it's nice, and update your map immediately.
• Extinction Learning: This is like learning that a dog you used to fear is now friendly. You have to unlearn an old fear and update the map to say, "The danger is gone."

2. The Anxiety Connection

The researchers found a strong link between how fast you update your map and how anxious you feel.

• The Finding: People who were slower to realize that a previously scary situation was actually safe (slow extinction learning) tended to have higher anxiety.
• The Analogy: Think of anxiety as a car alarm that won't turn off. The study found that people with anxiety often have a "stuck" alarm that takes a long time to realize the car is actually safe and the alarm should stop.

3. The Genetics Twist (The Big Surprise)

This is where the story gets interesting. Usually, when we find a trait linked to a disorder, we hope to find it in our DNA. The researchers checked if these learning speeds were "family traits" passed down from parents.

• Safety Learning: This was somewhat heritable. It's like having a genetic tendency to be naturally cautious or curious about new things.
• Extinction Learning (The "Stuck Alarm"): Surprisingly, this was not heritable. Even though it was linked to anxiety, your genes didn't seem to dictate how fast you could unlearn a fear.
• The Result: Because the "stuck alarm" (extinction learning) wasn't passed down genetically, the researchers concluded that it is not a genetic "endophenotype" (a biological fingerprint) for anxiety. You can't point to a specific gene and say, "This causes the slow unlearning."

The Bottom Line

The study confirmed that the way our brains handle unlearning fear is a real, measurable part of anxiety. However, it also showed that this specific struggle isn't something you inherit from your parents' DNA.

While we couldn't find the "anxiety genes" for this specific learning style, the study did prove that we can measure these learning processes using math and computer models, and that these measurements have a biological basis worth studying further. Essentially, we found a clear mechanical link between how we learn and how anxious we feel, even if that link isn't written in our genetic code.

Technical Summary

Technical Summary: The Heritability of Reinforcement Learning Parameters and Their Association with Anxiety

Problem Statement
Anxiety disorders are characterized by impaired learning regarding safety and the extinction of previously dangerous stimuli. These deficits are recognized as robust, heritable features of the disorder and are proposed as potential endophenotypes. While recent studies have established associations between the computational mechanisms underpinning these learning processes and anxiety severity, it remains unclear whether these specific learning parameters are themselves heritable or if they share a genetic basis with anxiety. This study addresses the need to replicate prior findings in a substantially larger sample and to rigorously test the heritability of these computational parameters.

Methodology
The authors conducted a pre-registered replication study utilizing a sample of 925 twins, representing a tenfold increase in sample size compared to previous investigations. The study employed computational modeling to estimate individual learning rates for two distinct processes: safety learning (learning that novel stimuli are safe) and extinction learning (learning that previously dangerous stimuli are now safe). The researchers analyzed the association between these learning rates and anxiety severity. Furthermore, they estimated the heritability ($h^2$) of both learning parameters and investigated the genetic overlap between the learning rates and anxiety severity.

Key Results
The study yielded a mixed pattern of replication and novel findings:

• Association with Anxiety: Extinction learning rates were significantly associated with anxiety severity (replication estimate $B = -0.14$, $BF_{01} = 1189.67$), confirming a robust cognitive-behavioral marker. In contrast, safety learning rates showed no significant association with anxiety severity.
• Heritability: Safety learning rates demonstrated modest heritability ($h^2_{safety} = 0.16$). Conversely, extinction learning rates were found to be non-heritable in this sample.
• Genetic Overlap: Due to the lack of heritability in extinction learning and the lack of association in safety learning, the study was unable to identify a genetic overlap between anxiety and either of the learning rates.

Significance and Claims
The paper concludes that neither safety nor extinction learning rates function as anxiety endophenotypes, given the absence of shared genetic architecture with the disorder. However, the study makes two distinct contributions to the field:

1. It confirms a specific cognitive-behavioral mechanism (extinction learning rates) that underpins a robust marker of anxiety severity, reinforcing the validity of computational modeling in understanding anxiety pathology.
2. It demonstrates the heritability of a computationally modeled learning parameter (safety learning), which represents a critical step toward establishing the biological basis of specific learning processes, even if those specific parameters do not directly serve as genetic markers for anxiety itself.