Annual herb, ascending or erect, attaining c. 1.25 m or more in cultivation. Stem stout, usually much-branched, it and the branches angular, glabrous or furnished in the upper parts with sparse (or denser in the inflorescence), ± crisped hairs. Leaves glabrous, or thinly pilose on the lower surface of the primary venation, green or purplish-suffused, very variable in size, long-(up to c. 8 cm) petiolate, the lamina broadly ovate, rhomboid-ovate or broadly elliptic to lanceolate-oblong, emarginate to obtuse or acute at the apex, at the base shortly cuneate to attenuate, decurrent along the petiole. Flowers green to crimson in ± globose clusters c. 4-25 mm in diameter, all or only the lower axillary and distant, the upper sometimes without subtending leaves and increasingly approximate to form a thick terminal spike of variable length, male and female flowers intermixed. Bracts and bracteoles broadly or deltoid-ovate, bracteoles subequalling or shorter than the perianth, pale-membranous, broadest near the base and narrowed upwards to the green midrib, which is excurrent to form a long, pale-tipped awn usually at least half as long as the basal portion and not rarely equalling it. Perianth segments 3, 3-5 mm long, elliptic or oblong-elliptic, narrowed above, pale-membranous, the green midrib excurrent into a long, pale-tipped awn; female flowers with the perianth segments slightly accrescent in fruit. Stigmas 3, erect or recurved, c. 2 mm. Capsule ovoid-urceolate with a short neck below the style-base, 2.25-2.75 mm, circumscissile, membranous, obscurely wrinkled. Seed 1-1.5 mm, black or brown, shining, very faintly reticulate, lenticular.

Type: Linnean specimen 117/7 (LINN, holotype!).

Distribution: Asia from India to China and Japan in the north and Indonesia in the south; also in New Guinea and New Hebrides and smaller Pacific Island groups (Fiji etc.). Introduced and/or cultivated in Africa, West Indies etc. Only habitat in Pakistan noted as “fields”.

I do not find the infraspecific divisions of this species devised by Aellen (in Hegi, Illust. Fl. Mitteleuropa ed. 2, Band 3/2 Lief. l:494-496) to be practicable. All three “subspecies” there recognised are sympatric, and all manner of difficulties arise in attempting to apply the system to material at hand. For example, two of the W. Pakistan specimens seen (Rahman s.n. and Stewart 15372) are identical in every character (including leaf shape, which is that of the type of Amaranthus mangostanus L.) except that one has developed a terminal spike (on the main stem only) but the other has not. One finds, too, on the same sheet as part of the same gathering, specimens with terminal spikes but one with leaves shaped as in the type of Amaranthus mangostanus and another with leaves as in the types of Amaranthus tricolor L. and Amaranthus melancholicus L.; similar leaf variation occurs in a single gathering of plants with axillary inflorescences only. Thus, since the significant characters alleged to separate the subspecies are to be found in all combinations, I prefer to regard Amaranthus tricolor as a polymorphic species with no attempt to subdivide. The only alternative would be to name the Pakistani specimens mentioned above as belonging to different subspecies, which would be absurd; had a large branch only been gathered of the specimen with a spike on the main stem, the two would have been virtually identical.