The most complete and definitive classification of cultivated Sorghum Sorghum bicolor (Linn.). Moench is that of Snowden (1936). All classifications since that time have been modifications or adaptations of the Snowden system. The "work groups" used by sorghum workers in India to classify the World Sorghum Collection (Murty et al., 1967) were based on it, as well as the system suggested by Jakushevsky (1969). In the classification proposed in this document, we freely admit our debt to the monumental and exhaustive studies of J.D. Snowden. On the other hand, people working with the crop in applied fields have found the Snowden classification extremely difficult to use. There are far too many names to memorize: 31 species, 158 varieties, and 523 forms for a total of 712 taxa. A vast array of character combinations is required to separate so many units. Many sorghums are not recognizable on sight and must be keyed out. Some of the taxa are based on very few specimens and some appear to be hybrid derivatives. One can often find several taxa in a single segregating F2 population, and new ones are produced in breeding programs. The geographic distribution of some taxa is so widely scattered or disjunct as to cast at least some doubt on the validity of implied relationships.

The difficulties of Snowden's classification are more fundamental than the over-fine partitioning of the variability. A hierarchical system was used with more categories than the biological facts will support. There exists a vast amount of variation within a category. This lends itself better to a parallel arrangement than one of hierarchies.


Linnaeus in 1753 described a Sorghum collection under the name Holcus: he delineated several species of Holcus, some which have been moved to the tribe Avenae, where the generic name Holcus now belongs. Adanson used the name Sorghum as an alternative for Holcus. In 1974 Moench distinguished the genus Sorghum from the genus Holcus (Celarier 1959; Clayton 1961).


Sorghum belongs to the tribe Andropogoneae. This tribe has a centre of variability in the Katanga (Congo) region of tropical Africa. Hartley (1958) suggested an Indo-Malaysian origin, there being many species in West India and South Indonesia. He considered that the tribe has a long evolutionary history in the Eastern hemisphere, but that it must have spread into tropical Africa at an early date.

Celarier (1956) regarded the Andropogoneae as being of more recent origin. Two of the primitive genera are Miscanthus and Miscanthidium. The former occurs in South East Asia and the Pacific islands, while Miscanthidium is found in tropical Africa. Celarier considered that Miscanthus was derived from Miscanthidium, tending to indicate Africa as the primary origin of the tribe.

The Sorghastrae is one of the 16 subtribes of the Andropogoneae (Stapf 1917). Garber (1950, 1954) Garber and Snyder (1951) split the subtribe into two main genera, and then subdivided the genus Sorghum into six subgenera. This classification is shown below.

Group Sorghastrae

  1. Genus Sorghum (Pers):
    1. Subgenus Eu-sorghum (Stapf), which is the same as Snowden's section Eu-sorghum;
    2. Subgenus Chaetosorghum;
    3. Subgenus Heterosorghum;
    4. Subgenus Sorghastrum (Nash);
    5. Subgenus Para-sorghum (Snowden);
    6. Subgenus Stiposorghum.
  2. Genus Cleistachne Bentham.

There are sharp distinctions between many of Garber's subgenera, which are now referred to as 'sections'. The term 'Eusorghum' has been discarded (de Wet 1978).

Celarier (1959) accepted much of this classification, but retained Sorghastrum as a separate genus and doubted whether the separation of Para- and Stiposorghum as distinct subgenera was justified.


The simplified classification

In 1972, Harlan and de Wet published a simplified classification of sorghum which has proved to be of real practical utility for sorghum workers. Instead of wading through the 31 categories in Snowden's key for cultivated sorghums, or attempting to chossed one of the 59 compartments of Murty and his co-workers (1967), a simple choice is now available. By looking at grain shape, glumes, and panicle, any cultivated sorghum can be assigned to one of five types, or to one of ten intermediate positions between any two of those types. Indeed, it may be possible to locate a position involving three of the five types. The system is simple, and it works: any cultivated sorghum can be assigned swiftly and consistently to one of the 15 groups. This system provides a good method for classifying cultivated sorghums, and would seem to be the categorical classification system of the future.

Harlan and de Wet partitioned the primary gene pool, Sorghum bicolor (Linn.) Moench, into the following races.

Cultivated races:

S. bicolor ssp bicolor
  1. Basic races:

    • Race (1) bicolor (B)
    • Race (2) guinea (G)
    • Race (3) caudatum (C)
    • Race (4) kafir (K)
    • Race (5) durra (D)
  2. Intermediate races: (all combinations of basic races)

    • Race (6) guinea-bicolor (GB)
    • Race (7) caudatum-bicolor (CB)
    • Race (8) kafir-bicolor (KB)
    • Race (9) durra-bicolor (DB)
    • Race (10) guinea-caudatum (GC)
    • Race (11) guinea-kafir (GK)
    • Race (12) guinea-durra (GD)
    • Race (13) kafir-caudatum (KC)
    • Race (14) durra-caudatum (DC)
    • Race (15) kafir-durra (KD)

  3. Spontaneous races: S. bicolor ssp arundinaceum.

    • Race (1) arundinaceum
    • Race (2) aethiopicum
    • Race (3) virgatum
    • Race (4) verticilliflorum
    • Race (5) propinquum
    • Race (6) shattercane

The system is clear and simple. Practically all of the variation in cultivated sorghum can be accounted for by the five basic races and their intermediate combinations. The races are, for the most part, easily identifiable by spikelet morphology alone. Intermediate races involving guinea, for example, have glumes that open partially and seeds that twist noticeably, but not as much as in pure guinea. Intermediate races involving caudatum have asymmetrical seeds, but the character is not as fully expressed as in pure caudatum. Other intermediate combinations can be recognised in a similar manner. The method is so sensitive that even three-way and possibly four-way combinations can be recognised, but these are usually products of modern plant breeding and not part of the variation of indigenous varieties. If they occur in significant numbers, they could be best be treated as sub-races of the main races.

The chief advantage of the system is its simplicity. One need only look at the spikelets and ask a few questions.

The degree of expression of these characteristics and their combinations determine the race, and the most part without equivocation. Identification can be made easily in the field or in the laboratory from head or even spikelet specimens.