Can Brain Connectivity Distinguish LSD, MDMA, and d-Amphetamine in One Person? What a 25-Volunteer fMRI Trial Shows

Quick answer: An experimental brain-imaging study found both shared and different average resting-state functional connectivity patterns after LSD, d-amphetamine, and MDMA. It did not validate a brain scan that identifies which drug a particular person took, measures a drug concentration, determines legal intoxication, or proves impairment during a specific incident. The most notable comparison was that LSD reduced default-mode network integrity relative to placebo, whereas that particular placebo contrast was not significant for d-amphetamine or MDMA. All three drugs reduced integrity in some other networks. These are group-level functional MRI findings, not a forensic toxicology classification test.

The primary experiment was reported by Avram, Fortea, Wollner, and colleagues in Molecular Psychiatry, published online in 2024 and in volume 30 in 2025: Large-scale brain connectivity changes following the administration of lysergic acid diethylamide, d-amphetamine, and 3,4-methylenedioxyamphetamine. The investigators reanalyzed imaging data from a randomized, double-blind, placebo-controlled, four-condition crossover clinical trial. Their neuroimaging sample comprised 25 healthy adults, not four independent groups of 25 people.

What forensic question can this study actually answer?

The scientifically supportable question is whether these three compounds produced different patterns in calculated correlations among resting brain-network signals under controlled experimental conditions. The answer is yes, for selected networks and contrasts. A separate, much stronger question is whether one can look at an unknown person’s scan and identify LSD rather than MDMA or d-amphetamine with a validated error rate. The experiment did not test that diagnostic problem. It did not recruit a separate blinded classification cohort, derive an individual decision cutoff, report drug-identification sensitivity or specificity, or perform a case-level prediction analysis.

That distinction matters whenever an expert is asked to translate an attractive group difference into an assertion about a defendant, driver, patient, witness, or decedent. A contrast significant across repeated laboratory sessions can coexist with broad overlap among individual participants. Statistical significance of a group effect does not supply the probability that any particular person took a given substance.

What substances and doses were administered?

The four crossover conditions were placebo, 0.1 mg lysergic acid diethylamide (LSD), 40 mg d-amphetamine, and 125 mg 3,4-methylenedioxymethamphetamine (MDMA). They were administered on separate study occasions in a randomized, counterbalanced order. The compounds are chemically and pharmacologically different: LSD is a classical psychedelic, d-amphetamine is a stimulant, and MDMA has both stimulant-related and prominent serotonergic actions. The study does not justify calling d-amphetamine methamphetamine, calling MDMA simply d-amphetamine, or treating these experimental doses as equivalent on a milligram-for-milligram potency scale.

The trial originally involved 28 volunteers. The brain-connectivity analysis retained 25 with suitable data, including 12 women, with a reported mean age of 28.2 years and standard deviation of 4.35 years. Each participant contributed observations across conditions. Thus the primary biological denominator is 25 distinct individuals. Network nodes, time points, image voxels, network pairs, and repeated sessions increase the number of measurements; they do not increase the number of independent people.

Study design and what was physically measured

A 3-tesla magnetic-resonance instrument acquired structural and resting-state functional MRI. The researchers processed spontaneous blood-oxygen-level-dependent (BOLD) time-series signals, adjusted for nuisance signals and movement, and evaluated statistical correspondence within and between predefined cortical networks. The resulting values describe synchronized fluctuations in a hemodynamic imaging signal. They are not physical chemical assays of LSD, amphetamine, or MDMA in the participant’s blood.

The publication also discusses pulse, blood pressure, tympanic temperature, head movement, and a questionnaire about altered subjective states. Those observations served as physiological or experiential context. Blood pressure is not a blood specimen; a magnetic-resonance signal is not a mass-spectrometric concentration; and a retrospective subjective questionnaire is not an instrumented driving trial. No venous whole-blood, separated serum, separated plasma, urine, oral-fluid, or vitreous drug-concentration series was quantified as an endpoint of the reported imaging comparison.

Evidence table 1: population, interventions, and measurement classes

The trial used the familiar labels visual (VIS), frontoparietal (FPN), default-mode (DMN), auditory-sensorimotor (ASM), dorsal attention (DAN), and salience (SAL). A candidate limbic network was excluded from detailed analysis because of concerns about signal-to-noise ratio and reproducibility. Even that seemingly technical decision changes the set of conclusions available for an individual interpretation: this is a prespecified collection of processed network metrics, not a picture of every biological process in the brain.

What is network integrity?

In this article, network integrity means the statistical consistency of resting-state activity within a network. The researchers used a dual-regression method and compared network-derived parameter estimates across the four conditions. Lower within-network integrity is a change in the correlations of the measured BOLD signals. It does not literally mean a network is anatomically damaged, that neurons are failing to connect, that someone cannot attend to a hazard, or that an individual necessarily experienced a clinically meaningful deficit.

Across all four experimental conditions, the authors reported false-discovery-rate-adjusted significant differences for the visual, frontoparietal, default-mode, and auditory-sensorimotor networks. The salience and dorsal-attention integrity analyses were not reported as significant omnibus network-integrity effects. An omnibus result asks whether at least some conditions differ; it does not by itself indicate which active drug differs from which other drug.

Evidence table 2: actual significant and nonsignificant network statistics

The significant p values represent inferential evidence about modeled condition effects under the chosen preprocessing and correction procedures. The nonsignificant values are equally important. An adjusted p value of .06 is not a positive finding merely because it is close to .05, and a head-motion p value of .15 cannot guarantee motion affected no participant or voxel.

Why was the default-mode network result distinctive?

The default-mode network is often examined in studies of internally directed thought and the shifting balance among brain systems. The observed placebo contrast for DMN integrity was significant for LSD, but not for d-amphetamine or MDMA. The study therefore supports a narrower statement than the proposition that an LSD scan is always recognizably different. It supports a group-level post-hoc imaging contrast in a defined network, using the paper’s pipeline and tested doses.

A useful forensic translation is to separate a feature that differs on average from a diagnostic feature that classifies an individual. There was no individual DMN-integrity threshold, no cross-validated confusion matrix, and no estimate of the likelihood of a false attribution in a person with anxiety, sleep deprivation, medication exposure, neurological illness, or mixed-drug use. None of those alternative populations was established as diagnostically separable in this trial.

What did LSD, MDMA, and d-amphetamine share?

Visual-network integrity declined under all three active conditions relative to placebo, as did frontoparietal-network integrity. Across the broader seed-based analyses, each drug also produced changes in connections among visual and higher-order networks. The overlap is scientifically instructive precisely because the agents have different leading pharmacological mechanisms. A common brain-imaging effect should not be inverted into the claim that all three agents must cause the same subjective experience, loss of motor control, or legal impairment.

The paper also found similarities between d-amphetamine and MDMA in the auditory-sensorimotor and salience patterns, which were more closely aligned than the investigators originally expected. Their discussion offers pharmacological hypotheses for the overlap, including dopaminergic and noradrenergic contributions. The network imaging data did not directly measure synaptic release of these transmitters in living human brain tissue, however. Those proposed mechanisms remain interpretations rather than simultaneously sampled neurochemical facts.

What is segregation between networks?

Segregation here describes the statistical relationship between the time-series signals assigned to different networks. Researchers often speak of increased connectivity between networks, which in this study corresponds to less segregation. This is a property of correlations and signal processing; it is not a direct physical count of axons and does not tell us that an individual has lost a specific psychological capacity.

LSD generally produced more extensive increases in between-network connectivity than d-amphetamine or MDMA. The paper identified especially pronounced changes involving transmodal networks, such as the default-mode and frontoparietal systems, and regions associated with sensory processing. But the direction and magnitude depended on the particular network pair and comparison. Some drug contrasts increased rather than decreased segregation. The study explicitly reported a nonsignificant corrected result for the auditory-sensorimotor/default-mode connection. Thus the phrase ‘more connected brain’ would erase important spatial and statistical differences.

What did the regional seed-based imaging show?

A seed-based analysis uses a network as a statistical starting point and maps how its time series corresponds to signals elsewhere in the brain. The original publication mapped increases and decreases across particular regions under each active compound compared with placebo. The investigators applied voxel-level and cluster-level statistical controls, then displayed spatial maps. These maps show where group contrasts survived the chosen familywise-error procedure. They are not heat maps of drug molecules, brain temperature, neuronal cell death, or measured individual behavioral errors.

For several networks, the strength of cross-network signal association increased after the active compounds. LSD often showed more widespread contrasts; d-amphetamine and MDMA showed overlaps and differences. Crucially, a printed colored patch from a statistical map cannot be assigned a case-specific likelihood of taking a particular drug without a separately trained and validated classifier. Even a striking figure is not the functional equivalent of a confirmatory chromatographic identification.

How did the global connectivity analysis differ?

The researchers also calculated degree centrality, a whole-brain graph-related measure of how strongly a voxel’s BOLD signal is connected to other voxels under a specified threshold and calculation method. All three substances were associated with decreases in visual-area degree centrality. Relative to placebo, LSD showed increases in selected basal ganglia and thalamic regions, whereas the corresponding thalamic increase was not observed for the amphetamines in the same comparison.

This could help generate neurobiological hypotheses. It does not prove a neurochemical mechanism, a cause of a hallucination, or a drug-specific marker for toxicology screening. Brain regions also participate in many ordinary behaviors and physiological processes. An imaging difference in one network metric is not a substance-exclusive fingerprint. The authors themselves found that some analytical choices, including global-signal regression, changed features of their results.

Why do receptor correlations require caution?

The article compared some fMRI connectivity maps with an existing population-derived map of serotonin 2A receptor density. LSD-related patterns showed stronger spatial associations with this reference map across several networks. That is biologically plausible and potentially useful for mechanistic research. However, the investigators did not obtain simultaneous receptor-binding scans for every subject under every drug condition. The receptor map and the changing connectivity signals were drawn from different sources.

Consequently, spatial correspondence is not a measurement of an individual’s occupied receptors or a demonstration that receptor activation caused that person’s observed behavior. Causation would require stronger evidence, such as appropriately designed pharmacological blocking or individual receptor-occupancy work combined with physiological and behavioral outcomes. This particular paper should not be stretched into that conclusion.

Were physiological variables or head movement considered?

Yes. Neuroimaging signals can be affected by head movement and changes in circulation or respiration. The researchers used correction procedures and assessed associations with framewise displacement and measured physiological changes. The mean displacement comparison among conditions was not significant, with F(3,72) = 1.80 and p = .15. They also acknowledged that residual physiological and movement effects could remain, particularly where associations changed after adding covariates.

The difference between ‘we controlled for a potential confound’ and ‘we eliminated the confound’ is material. A subject who moves during a scan can produce misleading connectivity estimates. A drug that changes heart rate or circulation can influence the hemodynamic signal without a one-to-one neuronal counterpart. The publication did not establish a correction procedure that makes a single person’s scan a certified drug-identification instrument.

What about subjective experience and impairment?

The original trial collected altered-state questionnaire information, and this analysis explored correlations of connectivity changes with reported subjective experiences. In the paper, some LSD- and MDMA-related imaging measures showed relationships with selected self-reported experiences. For d-amphetamine, fewer such associations appeared, mainly in network integrity. The researchers treated those findings as exploratory associations and discussed the potential impact of processing choices and covariates.

Even a reliable questionnaire finding would not establish an individual driver’s impairment at a different time or the cause of a crash. Subjective effects, visual-network correlations, divided attention, emergency braking, and lane maintenance are different endpoints. The Avram analysis did not measure standardized simulator driving, roadside field sobriety tests, accident reconstruction variables, or individualized psychomotor performance thresholds.

Nor can a particular self-report be reverse-engineered from an fMRI group average. A person can describe an experience under a substance without every imaging contrast being significant; conversely, group-average connectivity can change without establishing a clinically meaningful deficit for each participant.

What is missing for a forensic toxicology comparison?

Drug testing ordinarily begins with the exact collected specimen, collection time, chain of custody, target analyte, identification method, calibration, and uncertainty. If the question is recent exposure, the decision to test venous whole blood, serum, plasma, urine, oral fluid, or another matrix changes what can reasonably be inferred. If the question is impairment or causation, biological evidence must be connected to a separate behavioral, clinical, or incident record. These principles are explained in the site’s biological matrices guide and confirmatory drug-testing guide.

In this imaging study, no validated quantitative forensic whole-blood concentrations or specimen-specific pharmacokinetic curves were reported for the three study conditions. The researchers administered known laboratory doses, but there was no analytical link between each participant’s measured parent-drug blood concentration and the network estimates. There was also no mixed-drug condition designed to classify real-world polysubstance exposure, no laboratory intoxication decision rule, and no error-rate evaluation against blind unknown samples. That absence should be reported as ‘not measured’ or ‘not validated,’ not as a negative finding with an invented p value.

How should an attorney or expert handle an fMRI claim?

First, identify the proposition. Is an expert saying the drug can affect group-average network correlations, which this experiment supports for defined contrasts? Or is the expert claiming the scan proves someone ingested LSD rather than MDMA, which this study did not validate? The evidentiary burden differs sharply. A responsible source comparison identifies the specific participants, dose, protocol, magnetic-resonance preprocessing choices, network definition, test statistic, adjustment for multiple comparisons, and limits of the observation.

Second, preserve the null findings and alternative explanations. The nonsignificant auditory-sensorimotor/default-mode comparison and head-motion comparison belong in the analysis alongside significant visual and frontoparietal results. Different pre-processing choices and non-equivalent doses can affect apparent separations. An expert should not present an uncorrected regional finding as though the authors had validated a robust stand-alone diagnostic.

Third, distinguish the source of evidence. An original 25-person controlled fMRI crossover experiment is a credible experimental basis for the specific research question it tested. It is not a substitute for a qualified toxicology laboratory’s validated drug identification or a fact-sensitive assessment of a person’s cognitive or motor performance. The therapeutic drug-monitoring guide further illustrates why specimen timing and biological concentration answer different questions from an administered dose or an indirect physiological observation.

Seven frequently asked forensic questions

1. Can fMRI tell whether a particular person used LSD, MDMA, or d-amphetamine?

Not on the validation presented here. The researchers compared repeated condition averages in 25 healthy adults and reported selected significant networks. They did not derive or externally test an individual drug-identification classifier with sensitivity, specificity, or false-positive rates.

2. Did LSD change the default-mode network differently from the other two drugs?

In the reported placebo comparisons, LSD significantly reduced default-mode integrity whereas the corresponding d-amphetamine and MDMA placebo contrasts were not significant. This is a defined group-level observation; it is not a necessary condition for LSD exposure in every individual.

3. Did the three compounds affect all brain networks?

No. The investigators found significant omnibus integrity differences for four of six analyzed networks. Salience and dorsal attention did not show significant integrity omnibus effects, while other measures involving these networks sometimes changed. A network’s statistical result depends on the defined metric.

4. Was MDMA simply equivalent to d-amphetamine?

No. Their network effects overlapped in some analyses but differed in others, and the drugs have different pharmacological actions. Similar imaging contrasts are not therapeutic, toxicological, experiential, or psychomotor equivalence.

5. Did the scientists actually measure LSD or amphetamine in whole blood?

Not as a quantified outcome in the reported network comparison. They administered known experimental doses and acquired fMRI and physiological measures. An administered milligram dose cannot be converted into an individual’s venous whole-blood concentration without appropriate measurements and assumptions.

6. Does a significant imaging p value prove the person was legally impaired?

No. It indicates evidence of a modeled group condition difference under the specified statistical test. This article did not validate a legal impairment threshold, driving task, roadside assessment, or case-level causal determination.

7. Can a nonsignificant scan comparison prove two drugs have the same effects?

No. Lack of significance may reflect overlap, noise, sample size, or analytical choices. Demonstrating equivalence requires an equivalence or noninferiority design with defined margins and sufficient precision, not merely a p value above .05.

Bottom line: network patterns are not a chemical fingerprint

The strongest conclusion is measured and useful: in a 25-participant repeated-measures fMRI comparison, LSD, d-amphetamine, and MDMA showed both overlapping and distinctive brain-network signals under controlled dosing. LSD uniquely showed a significant default-mode integrity reduction versus placebo, while all three lowered visual and frontoparietal integrity. Some analyzed contrasts were nonsignificant, and preprocessing, participant selection, dose differences, and physiological effects limit individual interpretation. The original Avram and colleagues publication remains the authoritative primary reference for those comparisons: full original methods, figures, results, and discussion.

No part of those findings replaces a properly validated drug assay, provides a known person’s blood concentration or dosing time, or establishes the reason for a particular driving decision or injury. Forensic analysis must maintain the boundary between an experimentally observed group imaging metric and the separate evidence needed to evaluate an individual event.