Skip to content
PaperFren

How does meth dependence change learning and choices?

People with methamphetamine dependence rely on simpler, short-term decision strategies and show lower expectations of non-drug rewards compared to healthy individuals.

Source

Altered Statistical Learning and Decision-Making in Methamphetamine Dependence: Evidence from a Two-Armed Bandit Task

Harlé KM, Zhang S, Schiff M, et al. · Frontiers in psychology · 2015

doi.org/10.3389/fpsyg.2015.01910Read the full paper ↗31 citationscc by

What they did

The researchers compared 16 sober individuals with methamphetamine dependence and 16 matched healthy controls as they played 20 rounds of a two-armed bandit game. During the 16 trials of each round, participants chose between two options with fixed but unknown reward rates to earn points for cash. Researchers analyzed their choices using computational modeling to separate learning biases from decision strategies, and examined brain structure using magnetic resonance imaging.

What they found

While both groups earned similar overall scores, 50% of the methamphetamine-dependent participants relied on a simple win-stay/lose-shift strategy rather than a more complex learning strategy. Those who did use a learning strategy showed a lower reward maximization bias (1.58 compared to 6.99 for controls) and started with a lower prior expectation of reward (0.11 compared to 0.40 for controls). Additionally, reduced gray matter volume in the thalamus was linked to lower reward maximization, which fully explained the difference between the groups.

The limits

What it doesn't show

The study utilized a small sample size of only 16 participants per group, which limits the generalizability of the findings. Because brain structural differences were measured at a single time point, the study cannot determine if lower thalamic volume caused the decision-making deficits or resulted from prolonged drug use. Finally, because the participants were currently enrolled in a 28-day treatment program, these findings might not apply to active users or those with other psychiatric conditions.

Key terms

two-armed bandit task
A classic psychological paradigm where a participant makes repeated choices between two options with unknown reward probabilities to study the trade-off between exploring and exploiting.
Dynamic Belief Model
A Bayesian mathematical model that assumes the environment's statistical properties can change unexpectedly at any time, explaining how individuals update their beliefs over time.
Softmax policy
A decision-making strategy where options are chosen probabilistically based on their estimated values, with higher-value options being more likely to be selected.
Win-Stay/Lose-Shift
A simple heuristic strategy where a player repeats a choice after a successful outcome and changes choices after an unsuccessful outcome.
voxel-based morphometry
A neuroimaging analysis technique that allows investigation of focal differences in brain anatomy, such as gray matter volume, across the brain.
thalamic lateral dorsal nucleus
A brain region within the thalamus that is part of the limbic system and involved in emotional processing and spatial navigation.

Flashcards

1 / 35

Want these cards to stick?

Save the deck to NoteFren and study it with spaced repetition.

Save these cards to NoteFren— study “How does meth dependence change learning and choices?” with spaced repetition

Quiz yourself

1 / 10

Why is a computational modeling approach considered superior to coarse behavioral measures like total earnings in clinical psychology studies?

Common questions

Why did the participants' overall earnings not differ despite using different strategies?

The game was short (16 trials per round) and the reward probabilities were relatively close, meaning that even suboptimal strategies like win-stay/lose-shift could perform reasonably well on average, masking the subtle cognitive differences until computational modeling was applied.

What is the difference between maximizing and probability matching in the Softmax strategy?

Maximizing means consistently choosing the option with the highest expected reward, while probability matching means choosing options in proportion to their estimated reward rates (e.g., choosing a 70% rewarding option only 70% of the time), which is less optimal.

Did the study look at active methamphetamine users?

No, the participants were sober individuals recruited from an inpatient treatment program and were regularly screened for drug use to ensure sobriety during the study.

How does lower prior expectation of reward affect decision-making in real life?

A lower expectation of reward (or pessimism) for non-drug activities can lead to anhedonia, where natural rewards like money, social connection, or hobbies feel less appealing, potentially making drug-seeking behavior more attractive.

More on Addiction