Can Prenatal THC Affect Rat Memory While Anxiety Tests Remain Nonsignificant? A 43-Offspring Study

Short answer: In a controlled rat experiment, prenatal exposure to delta-9-tetrahydrocannabinol (THC) was associated with poorer performance on selected adolescent spatial-memory, object-discrimination, and reversal-learning tasks, alongside altered hippocampal messenger-RNA expression. Several other comparisons—including measured sucrose preference, open-field center transitions, elevated-plus-maze anxiety indices, and some working-memory outcomes—were not statistically significant. These were animal behavioral and brain-tissue measurements, not assays of a child’s blood THC, a clinical diagnosis of cognitive injury, or proof of impairment in a particular human case.

The primary study is Castelli and colleagues, “Prenatal Exposure to Δ9-Tetrahydrocannabinol Affects Hippocampus-Related Cognitive Functions in the Adolescent Rat Offspring: Focus on Specific Markers of Neuroplasticity,” Pharmaceutics (2023), DOI 10.3390/pharmaceutics15020692. Its strongest contribution is its endpoint-specific contrast: the researchers observed changes in certain memory tasks without corresponding significant changes in every measure of movement, reward response, or anxiety-like behavior.

What forensic question can this experiment actually answer?

The precise question is whether THC exposure administered to pregnant rats can precede measurable, selective changes in behavioral memory tests and hippocampal gene-expression markers in male offspring. The authors did not measure a witness’s memory, evaluate driving, perform a human neuropsychological examination, or reconstruct a mother’s cannabis use in a litigated pregnancy. Their observations are relevant to biological plausibility, not to automatic translation of exposure into a person-specific legal conclusion.

In litigation involving alleged prenatal substance exposure, three propositions often become conflated: a compound was administered or encountered; a group difference was observed in a laboratory endpoint; and that compound caused an individual person’s specific impairment or injury. This paper directly informs the second proposition under its experimental conditions. The first was controlled through dosing of pregnant rats. The third would require much more evidence and cannot be supplied by the rat experiment alone.

Who was studied: pregnant dams, offspring, and separate task denominators

Researchers obtained 12 pregnant Wistar rats at gestational day four. The mothers received daily subcutaneous THC at 2 mg/kg or a vehicle from gestational days 5 through 20. The investigators then studied 43 adolescent male offspring, selecting one or two males per litter for each experimental cohort. Testing began around postnatal days 35–46. These 43 offspring were not all tested on every behavioral task, and the 12 mothers were not 43 independent prenatal exposures.

One offspring cohort underwent the open-field, elevated-plus-maze, and Barnes-maze procedures. Another underwent sucrose preference and the reward-motivated Can test. The published figures specify different denominators: the Can-test plots report nine controls and ten prenatal-THC rats; the Barnes-maze plots specify 12 rats per group; the sucrose-preference plot reports nine controls and eight exposed rats. Gene-expression testing involved a selected hippocampal-tissue subset, not all 43 animals.

Multiple offspring from a mother share the same prenatal environment. Because treatment was administered at the dam level, litter clustering matters when judging inferential independence. A nominal rat-level p value should not silently be treated as though every offspring were the product of a separate randomized pregnancy. The paper’s methods describe statistical comparisons across offspring but do not establish that litter-level correlation was fully addressed for every endpoint.

What did prenatal exposure consist of?

The administered material was THC resin described as more than 97% pure by the authors’ analytical characterization. The dose was formulated in an ethanol/Tween/saline preparation and given subcutaneously to pregnant dams. That is not smoked marijuana, edible cannabis, prescribed oral THC, or passive secondhand exposure. The 2 mg/kg figure refers to an administered maternal dose, not a physically measured fetal blood or brain concentration.

The authors described their regimen as intended to model relatively mild human THC consumption; that description is a study rationale, not a validated quantitative conversion to a pregnant person’s dose or drug level. The present report contains no direct paired measurements of THC in maternal venous whole blood, maternal serum, cord blood, neonatal blood, or adolescent offspring blood. It therefore cannot support an individual blood-concentration threshold, fetal dose estimate, or timing-of-use calculation.

Why were two different memory tasks used?

The Can test is a reward-motivated task. Water-restricted rats learned which of several cans contained a water reward, either from the can’s fixed location or from its distinctive appearance despite changing position. The researchers recorded proportions of correct first choices, reference-memory errors, repeat visits interpreted as working-memory errors, and whether an animal was active enough to explore. It is a controlled behavioral test, not a direct measurement of human episodic memory.

The Barnes maze uses a circular platform, visual room cues, and an escape location. Bright lighting and an elevated open platform motivate animals to find the escape box. The acquisition phase concerns learning an escape location; a later probe tests recall of that location; a reversal phase moves the escape location to assess adjustment to a changed rule. These different phases are essential: not all of them produced significant differences.

Can-test spatial memory: the strongest positive comparisons

In the fixed-position Can task, the exposed offspring had a lower proportion of correct first responses than controls on the second and third testing days. The authors reported post-hoc p = 0.0099 on day two and p = 0.0059 on day three. For reference-memory errors—first visits to unrewarded cans—the day-two comparison was p < 0.0001. Repeated-measures analyses also found a group-by-day interaction for correct responses (F(2,34) = 3.370, p = 0.0462) and for reference errors (F(2,34) = 19.13, p < 0.0001).

The distinction is not trivial: the study did not find a prenatal-THC group main effect on working-memory errors in this task (p = 0.4003) or on the reported activity score (p = 0.8782). That combination supports a selective behavioral interpretation better than the blanket statement that prenatal THC impaired every facet of cognition or made exposed rats unwilling to perform the task. The plotted values are proportions; the p values are not percentages of affected offspring.

Object discrimination: poorer first choices but not every error type

In the Can task’s simple-visual phase, a visually different rewarded can could move among otherwise identical cans. Prenatally exposed rats made fewer correct first choices overall (F(1,17) = 14.75, p = 0.0013) and more reference-memory errors (F(1,17) = 21.83, p = 0.0002). These group effects are relevant to object discrimination and reference memory within this specific reward-motivated paradigm.

Here again, some findings were null: the group main effect on working-memory errors was p = 0.6715, and the group effect on task-activity score was p > 0.9999. Non-significance does not prove identical performance, especially in small groups. It means these particular analyses did not detect the claimed group effect at the chosen threshold.

Barnes maze: retention and reversal differed; one acquisition criterion did not

During Barnes-maze acquisition, a repeated-measures analysis of escape latency reported a prenatal-treatment main effect (F(1,51) = 4.878, p = 0.0317). But a separate analysis asking how many days each rat took to reach the learning criterion did not reach significance: log-rank χ² = 2.608, p = 0.1064. The reported mean days to criterion were 1.458 for controls (95% CI 1.129–1.787) and 1.759 for prenatal-THC offspring (95% CI 1.427–2.090). The distinct outcomes should not be merged into an assertion that all acquisition metrics proved impairment.

For the later probe task, exposed rats took longer to make initial contact with the remembered target location (t = 2.324, df = 22, p = 0.0297). For reversal, after the escape location rotated 180 degrees, their latency to find the new escape was also longer (t = 2.921, df = 22, p = 0.0079). These were laboratory measures of rat maze behavior, not observed human driving mistakes, witness-identification errors, or clinical diagnostic scores.

Evidence table 1: measured behavioral endpoints and important nulls

Table reading: Each row is a particular endpoint from the original experiment; repeated trials and tested offspring must not be mistaken for independent pregnancies. The displayed p values are source-reported and do not measure the probability that an individual person’s injury was caused by THC.

Movement, exploration, anxiety-like behavior, and sucrose preference

Potential motor or motivational explanations were considered. In the open-field test, exposed rats traveled a greater total distance (U = 25, p = 0.0053), while center-transition counts did not significantly differ (p = 0.4248). On the elevated plus maze, percentage of time spent in open arms (p = 0.6188) and percentage of open-arm entries (p = 0.3824) were not significantly different, even though total arm entries were higher among exposed offspring (p = 0.0078).

Likewise, a sucrose-preference comparison did not detect a group difference (p = 0.1035; figure n = 9 controls and 8 exposed). These findings argue against simply relabeling every adverse memory-task observation as generalized lethargy, low reward motivation, or a uniform increase in anxiety. But they do not establish that all possible emotional, motivational, or sensory confounders have been eliminated; the specific tests and their measurement precision constrain that interpretation.

What was physically measured in hippocampal tissue?

The investigators dissected hippocampal tissue from a selected subset of adolescent male offspring and measured messenger-RNA abundance by quantitative reverse-transcription PCR (qRT-PCR). The assayed targets included the NMDA receptor subunit transcripts Grin1/NR1 and Grin2a/NR2A, the postsynaptic scaffolding transcript Dlg4/PSD95, the metabotropic glutamate receptor Grm5/mGluR5, Homer1, the cannabinoid receptor transcript Cnr1/CB1R, and Hint1/HINT1. The figure shows distinct tissue-subset observations rather than concentrations of THC in the tissue.

These are gene-expression measurements, not direct protein quantitation, synaptic-current recordings, counts of dead neurons, permanent DNA mutations, or clinically established brain damage. The paper reports significant differences in transcripts that participate in excitatory and endocannabinoid-related signaling. A change in a transcript is a molecular observation whose functional consequences cannot be read as a direct measure of a person’s ability to remember an event.

Evidence table 2: hippocampal gene-expression direction and tests

Table reading: The published t-test degrees of freedom of eight are consistent with five sampled animals per group for those comparisons. The report does not support treating the 43 offspring, the number of PCR wells, or the total RNA molecules as independent tissue biological replicates. The direction and p values describe transcript differences; they do not validate a molecular diagnostic test for an individual child’s impairment.

Why hippocampal RNA changes do not diagnose synaptic injury

The experimental interpretation is that prenatal THC exposure may shift the balance of genes relevant to excitatory neurotransmission and endocannabinoid feedback. NR1 and NR2A contribute different functional parts of NMDA-receptor signaling; PSD95 is associated with receptor organization; mGluR5 and Homer1 participate in glutamatergic and endocannabinoid signaling; CB1R is an important cannabinoid-system receptor; and HINT1 has regulatory roles. The direction of transcript changes was not uniform. For example, NR1 increased while NR2A and PSD95 decreased.

It would overstate the experiment to say that increased NR1 mRNA proves excessive neuronal firing, or that reduced PSD95 mRNA proves permanent neuronal loss. The investigators did not establish the corresponding human protein abundances or identify a threshold at which an individual person becomes cognitively impaired. Behavioral group differences and molecular group differences were observed in the experimental model; a quantitative mechanistic chain connecting a particular prenatal human exposure to a particular legal injury was not measured.

Why actual analytical specimens matter in a forensic interpretation

Three very different kinds of evidence must be kept apart here. First, the protocol documents a maternal administered mass per kilogram. Second, the team observed offspring laboratory behavior on specified tasks. Third, they measured offspring hippocampal RNA after sacrifice. None of these is a measured concentration of THC in the maternal or offspring forensic blood matrix.

Blood, serum, plasma, urine, oral fluid, hair, and tissue cannot be translated into each other without a validated method and the relevant collection history. Our guide to biological matrices explains why a result must stay tied to its specimen and method. This study does not include a validated confirmatory THC result from forensic venous whole blood or a urine metabolite test capable of dating prenatal use. Nor does it provide a breath THC value, an exposure-response curve, or a generalizable fetal-dose reconstruction.

Where a party relies on a purported drug concentration or screening result, method-specific confirmatory testing is a separate question from the animal behavioral experiments described here. The chain from collection to analysis also matters; see specimen collection and handling. These contextual pages are not substitutes for the complete source study or the original laboratory record in a case.

What the study cannot establish about a child, driver, or eyewitness

No pregnant human participants were enrolled. No infants or children were followed clinically. No adverse childhood diagnosis was assessed. No driver performed an on-road task. No eyewitness identified a suspect or recalled an incident. No clinician applied a human neuropsychological battery. Consequently the reported rat task statistics cannot be converted to a percentage decline in a child’s intelligence quotient, a human impairment score, a probability of negligent driving, or the reliability of eyewitness memory.

Equally, the lack of significant change in selected elevated-plus-maze or sucrose comparisons cannot establish that prenatal THC exposure is harmless. An animal endpoint may be insensitive to some harms while detecting another; different ages, routes, exposure magnitudes, sexes, species, or endpoints may behave differently. A valid account must preserve both the observed positive results and the genuine nulls without extending either further than this design allows.

What are the main statistical and experimental limitations?

Sample structure: the study began with 12 treated or control pregnant dams and measured 43 male offspring, with multiple offspring per litter possible. Litter-related dependence is a separate statistical issue from rat counts and task repetitions. Sex restriction: adolescent males were studied; a female-offspring or human sex-specific inference is not available from this dataset. Route: mothers received subcutaneous THC, a different absorption setting from human smoking, vaping, or oral ingestion. Timing: the study assessed adolescence after prenatal exposure, not the instantaneous impairment state at a human event. Multiplicity: many behavioral and transcript endpoints were analyzed, so individual p values should be interpreted in the context of the whole set of tests and the correction methods the authors reported.

Mechanism: hippocampal qRT-PCR detects transcript abundance; it does not directly determine receptor protein, electrophysiological function, irreversible brain pathology, or why a particular human may have a symptom. External validity: controlled Wistar-rat conditions reduce some experimental variability but do not eliminate maternal, litter, developmental, environmental, or species differences. Individualization: this publication does not supply a validated clinical diagnostic sensitivity, specificity, predictive value, or causal attribution algorithm.

Seven frequently asked forensic-science questions

1. Does this paper prove that prenatal THC causes memory loss in children?

No. It documents group differences in selected spatial and object-memory measures among adolescent male rat offspring under a controlled gestational injection regimen. It did not study children or supply human individual diagnostic criteria. A conclusion about a specific child would require independent human exposure, clinical, developmental, and alternative-cause evidence.

2. How many animals were actually tested?

Twelve pregnant dams were used and 43 male offspring entered separate experimental cohorts. The task-specific biological denominators varied: the Can-task plots use nine controls and ten exposed offspring, Barnes-maze plots report 12 per group, and the sucrose plot reports nine versus eight. Selected hippocampal RNA analyses involved smaller subsets. Trials and technical measurements are not additional independent animals.

3. Did THC impair working memory in every test?

No. Group main effects on the source’s Can-test working-memory-error measures were nonsignificant: p = 0.4003 in the spatial task and p = 0.6715 in the visual discrimination task. Reference-memory and correct-choice outcomes did show differences. It would be inaccurate to describe every memory domain as significantly impaired.

4. Did exposed rats have significantly more anxiety-like behavior?

The elevated-plus-maze open-arm time and entry percentages did not significantly differ, with p = 0.6188 and p = 0.3824 respectively. Total entries increased, which is an activity measure rather than proof of increased anxiety. An absence of significance in these tests is not proof of complete emotional equivalence.

5. Was THC directly measured in the offspring’s blood or brain?

No direct offspring blood or brain THC concentration is reported in this study. The prenatal exposure was a maternal administered injection. The dissected hippocampi were tested for selected mRNA markers by qRT-PCR, not for quantitative parent THC, blood metabolites, or forensic toxicology cutoffs.

6. Does a significant hippocampal NR1 or CB1R RNA result prove permanent injury?

No. Altered RNA abundance is a molecular endpoint. It is not a measurement of human neuronal death, receptor protein quantity, permanent DNA mutation, or lifetime functional disability. The experimental group differences should not be used as an individual medical diagnosis.

7. Can these findings determine impairment or responsibility in a particular DUI, injury, or eyewitness case?

No. The study contains no human driving, eyewitness, case-specific behavioral observations, or quantitative human THC matrix results. Its appropriate role is limited preclinical context. Case-specific conclusions require the actual analytical method and specimens, event timing, observed performance, medical and developmental history, competing explanations, and legally relevant proof.

Bottom line: specific rat endpoints, not a universal human impairment claim

Castelli et al. observed reduced performance on some adolescent male rat memory tasks after a controlled maternal prenatal THC regimen, while several working-memory, sucrose-response, and anxiety-like comparisons remained nonsignificant. They also measured directional changes in hippocampal synaptic-signaling transcripts. The scientific value is in that specific pattern—not in pretending the experiment measured a child’s blood THC, proved permanent neurological injury, or reconstructed an individual’s later impairment.

Original primary source: Castelli V, Lavanco G, Feo S, D’Amico C, Micale V, Kuchar M, Plescia F, Brancato A, Cannizzaro C. Pharmaceutics. 2023;15(2):692. DOI: 10.3390/pharmaceutics15020692.