A plant
through time.

A 2023 hop-genome study estimated the Cannabis–Humulus divergence at about 22.6 million years ago using a fossil-calibrated Bayesian molecular clock. Its 95% highest posterior density interval extended from approximately 15.7 to 28.8 million years. Other methods in the same study produced estimates around 16–17 million years. The spread matters: fossil calibration, sampled genes and statistical method influence the date inferred from living genomes.

Where the evidence stopsA molecular estimate is not a directly observed birth date. It does not identify the first cannabis individual, settle the plant’s precise birthplace, or establish where people first used it.

McPartland, Hegman and Long reviewed 155 Asian fossil-pollen studies in 2019. Their oldest Cannabis–Humulus pollen record consistent with cannabis was dated to about 19.6 million years ago in northwestern China. Because cannabis and hop pollen can be difficult to distinguish, they used accompanying steppe or forest pollen as ecological clues. Their proposed origin near Qinghai Lake also draws on the reconstructed history of steppe vegetation.

Where the evidence stopsPollen identification and ecological inference remain uncertain. This hypothesis does not establish a particular Himalayan valley as the plant’s origin, and it cannot reveal cultivation or the chemical composition of these ancient plants.

A 2021 genomic study modelled a domestication-associated divergence around 12,000 years before present, with a broad 95% interval of roughly 6,458–15,728 years. Separately, directly dated Cannabis fruits from Okinoshima in Chiba have a calibrated range reported as 8280–7660 BCE in a later archaeobotanical review. The genomic result is an inference from modern samples; the Japanese finds are physical remains, with possible cultivation discussed by their investigators.

Where the evidence stopsBP means years before 1950; BCE uses a calendar scale. The evidence does not identify the first grower or prove that every early occurrence represents cultivation, fibre production or intoxicating use.

Research on Haimenkou reports more than 800 Cannabis grains from contexts dated approximately 1650–400 BCE, most concentrated in a deposit from 1400–1100 BCE. The same study reviews a securely identified roll of hemp cloth from Shang-period Gaocheng Taixi, approximately 1600–1046 BCE. Seed measurements support discussion of an early crop with multiple possible uses; they do not directly measure ancient cannabinoid chemistry.

Where the evidence stopsThese are separate sites and contexts. They do not demonstrate identical practices across China, reveal the use of every grain, or validate every proposed earlier hemp textile.

A 2019 study analysed residues from braziers at Jirzankal, dated to about 500 BCE. Cannabinol, a degradation product associated with THC, supported the interpretation that cannabis was burned and its smoke inhaled during funerary rituals. The authors inferred relatively THC-rich material. This is among the earliest directly dated, chemically supported evidence for such ritual use; the study did not recover a measurable original THC percentage.

Where the evidence stopsChemical residues do not supply a complete ritual narrative or a modern potency comparison. This is not the first evidence of all cannabis use, nor evidence that the practice provided a medical benefit.

Brand and Zhao’s 2017 review places the Shen Nong Ben Cao Jing in the first or second century CE. Their examination of Chinese medical literature distinguishes references to seed fruits from other plant parts and explores how historical terminology complicates interpretation. The work’s traditional association with Shennong does not date its composition to the third millennium BCE. Its historical indications require interpretation within the medical concepts of their periods.

Where the evidence stopsA documented historical prescription is not modern clinical evidence. Legendary authorship should not become a dated discovery claim, and ancient plant-part names cannot always be mapped confidently onto current preparations.

Patricia Sauthoff’s 2024 study revisits earlier philological research, including Dominik Wujastyk’s identification of an eleventh-century reference in Vaṅgasena’s Cikitsāsārasaṃgraha to ingestible medicinal or intoxicating cannabis. The twelfth–thirteenth-century Ānandakanda offers an extended discussion involving cultivation, preparations and dangers. The study also examines ambiguous names and challenges some retrospective identifications. These texts support a medieval written history; they do not establish when cannabis first reached or was used in South Asia.

Where the evidence stopsEarlier bhaṅga references and identifications of soma are contested. Textual claims, mythology and prescriptions are not modern safety or efficacy findings, and no single manuscript establishes an Indian invention of cannabis use.

Harvard scholar Adam Bremer-McCollum’s 2025 research reflection examines fourteenth- and fifteenth-century Arabic sources by historians, jurists and poets. It dates al-Badrī’s two-part collection to the 1460s and discusses regional vocabulary alongside opposing evaluations. These are textual witnesses read through historical scholarship. Their different voices help document how people wrote and argued about cannabis; they are not population surveys or a chronology of the plant’s arrival.

Where the evidence stopsThese writings do not express a universal Islamic position, prove a first use, or establish that every region shared the same practices. Literary voices must remain distinct from direct observation.

The Mary Rose conservation team identifies its recovered anchor cable as hemp, with more than one hundred metres raised. The ship sank in 1545. NICAS identifies Scheurrak SO1’s surviving sailcloth as hemp fabric. The Dutch Cultural Heritage Agency describes that wreck as a merchant ship carrying Baltic wheat, lost shortly before 1600. These are documented material objects, rather than a general claim about every ship of the period.

Where the evidence stopsThese finds do not establish that all contemporary sails were hemp, identify where these fibres grew, or prove a maritime first. Scheurrak’s exact sinking date remains uncertain.

O’Shaughnessy’s 1843 journal publication identifies an earlier Calcutta report from 1839 and acknowledges Indian and Persian knowledge alongside his observations. The Indian Hemp Drugs Commission of 1893–94 subsequently assessed cultivation and consumption. Its report opposed total prohibition and favoured restriction through taxation, while considering harms and unintended effects. These are historical positions from a colonial inquiry, not the conclusions of a contemporary controlled clinical study.

Where the evidence stopsO’Shaughnessy did not discover that cannabis had medicinal uses. The Commission’s conclusions should be attributed to its historical setting and cannot establish that cannabis is harmless or effective as a modern treatment.

The United States Department of Agriculture’s Hemp for Victory is dated circa 1942 in the National Agricultural Library archive. The film documents cultivation, harvesting and the manufacture and military uses of hemp fibre and coarse cloth. It is an original record of a specific wartime campaign, providing evidence of what that campaign promoted and how it presented the crop to its intended agricultural audience.

Where the evidence stopsA wartime promotional film is not a neutral account of the plant’s entire history. It does not mark the invention of hemp fibre or imply unrestricted permission for all cannabis cultivation and use.

Their April 1964 paper in the Journal of the American Chemical Society describes isolation, structural identification and partial synthesis of an active constituent of hashish. This is a foundational THC milestone, not the discovery of the first cannabinoid or the completion of every subsequent stereochemical and synthetic question. The historical publication date is 1964; a later date on a digital publisher record reflects online availability.

Where the evidence stopsIdentifying a molecule does not prove that a plant preparation treats a disease. THC chemistry should not be attributed to hemp seed oil or protein, and this paper was not a clinical trial.

The NDPS Act was enacted on 16 September 1985 and came into force on 14 November. Section 2(iii) defines cannabis through charas, ganja and specified mixtures; seeds and leaves are excluded from ganja when unaccompanied by the tops. Section 14 allows government authorization, subject to conditions, for fibre, seed or horticulture. The 1961 Convention separately excludes cultivation exclusively for industrial fibre and seed or horticultural purposes from its application.

Where the evidence stopsThese provisions do not automatically authorize cultivation or establish blanket legality for bhang. This is historical legal context, not advice about a current activity; applicable permissions and rules depend on jurisdiction and circumstances.

Devane and colleagues used a radiolabelled cannabinoid compound to characterize binding.

Where the evidence stopsAnimal-tissue binding is not a patient trial, and the receptor had not yet been cloned.

Matsuda and colleagues reported a cannabinoid-responsive G protein-coupled receptor.

Where the evidence stopsThis was a cloning milestone, not discovery of the entire endocannabinoid system or proof of a treatment benefit.

The work identifies an endogenous ligand, rather than a constituent supplied by the plant.

Where the evidence stopsIt does not mean the body needs cannabis, or establish a benefit from consuming it.

The original study reported receptor cloning and expression in splenic macrophages.

Where the evidence stopsIts historical expression findings are not a complete modern map of CB2, and do not establish an immunity benefit.

Mechoulam and colleagues reported the intestinal-tissue finding in June; Sugiura and colleagues reported brain evidence in October.

Where the evidence stopsThese studies concern a cannabinoid-ligand role for an existing lipid. They do not map the complete signalling system or demonstrate clinical efficacy.

Over 14 weeks, convulsive seizures declined more with cannabidiol. Adverse events and withdrawals were more frequent.

Where the evidence stopsThis is evidence for a defined formulation and syndrome. It does not establish equivalent effects for seed oil, protein powder, every CBD preparation or all epilepsy.

The FDA’s original approval package for Epidiolex, dated 25 June 2018, covers seizures associated with Lennox–Gastaut syndrome or Dravet syndrome in patients aged two years and older at that time. The decision followed a product-specific assessment and included safety obligations, including attention to liver injury. This chapter records the initial 2018 authorization, rather than presenting its original indication and age range as the complete current prescribing label.

Where the evidence stopsThis approval does not cover hemp seed oil, protein, ordinary CBD products, whole cannabis or Mount Herbs goods. It is a U.S. regulatory milestone, not treatment guidance or a claim of universal safety.

The supplemental approval follows the original 2018 authorization for Dravet and Lennox–Gastaut syndromes.

Where the evidence stopsThis US decision is not an Indian authorization, universal cannabis approval or evidence for consumer seed products.

Decision 63/17 was adopted on 2 December 2020 by 27 votes to 25, with one abstention. It deleted cannabis and cannabis resin from Schedule IV while their Schedule I status remained. The WHO’s explanation emphasized that international controls continued. The decision therefore should not be described as removing cannabis from all international schedules, legalizing it worldwide, or approving every cannabis-derived product as a medicine.

Where the evidence stopsInternational rescheduling does not automatically rewrite every domestic law. This event establishes neither commercial permission for a particular product nor clinical efficacy for a particular preparation, condition or person.

Reference Material 8210 has non-certified reference values intended for laboratory controls and research.

Where the evidence stopsA reference material is not certification of a brand’s products, proof of purity or evidence of treatment benefit.

Among 147 evaluable participants, complete response was 24% with extract versus 8% with placebo. Sedation and dizziness were more common.

Where the evidence stopsComplete response meant no emesis and no rescue medication over 0–120 hours. It was not a tumour response or complete absence of nausea.

Lynch and colleagues’ 2025 study assembled a cannabis pangenome resource comprising 193 genomes from 144 biological samples: 181 new assemblies and 12 existing ones. It documented extensive genomic diversity. A divergent Tibetan assembly contributed evidence suggesting wild relatives may persist in Asia, while the authors called for wider Asian and historical sampling. This complicates a simple account in which all wild progenitors have certainly disappeared.

Where the evidence stopsOne divergent genome does not prove a birthplace or settle wild versus feral status. A pangenome is a research resource, not evidence of a product’s clinical benefits or a final answer to domestication.

Condition, formulation, comparator and follow-up must remain attached to every result. Later studies may fall outside an earlier review’s search cutoff.

Where the evidence stopsSelected evidence checked 3 October 2026; not an exhaustive or continuously updated review. Negative and uncertain results remain part of the story.