The Binary Structure of Ifá: How 256 Odù Are Generated and Read

An Odù is written as two columns standing side by side, each column holding four positions stacked vertically, and each position carrying either one mark or two. A position with one mark and a position with two marks are the only options, which makes each position a binary digit. Four positions per column gives sixteen possible columns, two columns gives sixteen times sixteen, and sixteen times sixteen is 256. That is the whole of the arithmetic, and everything else in this file is detail about how the marks are produced, how the figure is oriented, and what the structure does and does not entitle you to claim.

Bascom states the formula directly: each half of a figure takes one of sixteen basic forms "according to the formula 2^n for the number of permutations of heads and tails for a coin tossed four times in succession, with 2 being the number of possible alternatives (heads or tails) and n the number of tosses" 1 He then gives the combination: "a total of 16 x 16 or 256 derivative figures (odu) are possible" 1

The mark, and what it is made of

The unit is a mark drawn in wood powder on a tray, or a seed shell lying one way up rather than the other. It has two states and no third.

With ikin, the sixteen sacred palm nuts. The diviner holds all sixteen nuts in one hand and grabs at them with the other. Almost always some remain behind. The rule Abimbola gives is that if one nut remains he makes two marks, and if two nuts remain he makes one mark 2. Bascom gives the identical rule and flags what is strange about it: "If one nut remains, the diviner draws two short parallel lines in the divining powder on the tray; if two nuts remain, he makes a single line. In justification of this apparently arbitrary reversal, the diviners say only that this is how Ifa taught the Ifẹ̀ to do" 1

Note the reversal, because it matters for anyone trying to read the system as a clean numeral notation. One nut yields two marks; two nuts yield one mark. The mapping between the physical event and the recorded symbol is inverted, and no functional explanation for the inversion is recorded. The tradition's answer is that this is how it was taught.

The trial only counts when exactly one or two nuts remain. Bascom: "If none is left, or if more than two remain, or if the grasp is not secure and he feels that some are 'trying to slip out,' he beats them again and takes another grasp" 1 Abimbola gives the same rule 2. This is a rejection-sampling procedure. Outcomes outside the valid range are discarded and the draw is repeated, which is a real and specific piece of engineering: it converts a physical process with many possible outcomes into a clean two-state output.

That procedure is repeated until eight valid marks exist.

With ọ̀pẹ̀lẹ̀, the divining chain. Eight half seed shells, four on each side of a chain held at the middle. Each shell lands either concave side up or convex side up. Bascom: a half shell falling "open" with the concave inner surface upward "is equivalent to a single line on the tray; if it falls in the closed or 'inverted' position with the convex outer surface up, it is the equivalent of a double mark" 1 Abimbola describes the same correspondence in terms of the rough inner and smooth outer surfaces 2.

The chain produces all eight bits in one throw. The palm nuts require eight separate manipulations, each of which may need several attempts. Bascom states the contrast plainly: "One of the 256 figures can be selected by a single cast of the divining chain, whereas eight separate manipulations of the palm nuts are required" 1 The two methods are informationally identical. Bascom: "otherwise the two systems are identical. They employ the same set of figures with the same names and rank order, and the same verses" 1

Orientation, and why it is not decorative

The eight marks are made in a fixed order, and the two columns are not interchangeable.

Bascom: the eight marks "are arranged in two parallel columns of four marks each" and "the diviner works from right to left in marking the lines on the divining tray, in reading the figure, and in choosing between specific alternatives" 1 Abimbola describes the same: the diviner "makes either one or two marks in this way four times in two rows starting from the right to the left" 2

So the right column is filled first, top to bottom, then the left column, top to bottom.

The right column takes priority in naming. Bascom: "Each derivative figure is given a compound name based on the names of the patterns of the right and of the left side. The name of the right side precedes that of the left side" 1 He adds the interpretive gloss the tradition attaches: the right (ọ̀tún) "is regarded as male and 'more powerful' than the left (òsì), which is female; for this reason the name of the right half precedes that of the left" 1

This is not a notational nicety. Bascom is explicit that reversing the columns produces a different Odù with different content: the illustrated figure "is Ọ̀kànràn Ìrẹtẹ̀, and not Ìrẹtẹ̀ Ọ̀kànràn. Because the latter is another figure, with a different set of predictions and sacrifices, it is essential to differentiate between the two halves of the figure by orienting the divining tray and the two halves of the divining chain properly" 1

Abimbola confirms the same convention from the naming side, glossing a minor Odù: "Its actual name is Ogbégúndá, that is a combination of Ogbè on the right and Ògúndá on the left" 2

A practical consequence for readers of secondary sources: a great deal of writing on Ifá, including material produced by practitioners, presents Odù diagrams without stating which column is which. Since the ordered pair is what identifies the Odù, an unlabelled diagram is ambiguous between two different Odù unless the figure happens to be symmetric.

Notation used in this corpus

Sources differ in how they write the marks, so this corpus fixes one convention and states its mapping.

We write a column top to bottom as a string of four digits, using Bascom's own notation from his Table 3 1:

• 1 = a single mark (one stroke; ọ̀pẹ̀lẹ̀ shell concave side up)
• 2 = a double mark (two strokes; ọ̀pẹ̀lẹ̀ shell convex side up)

A full Odù is written RIGHT|LEFT, right column first, matching the naming order. So Èjì Ogbè is 1111|1111 and Ogbèyèkú is 1111|2222, that being Ogbè on the right and Ọ̀yẹ̀kú on the left.

To convert to conventional binary, map single mark to 1 and double mark to 0, reading the top position as the most significant bit. This is a choice made for this corpus, not something the tradition specifies, and the caution in the next section applies to it.

The sixteen columns, verified

Under Bascom's notation the sixteen named column patterns are these, and they are mathematically complete: all sixteen possible four-position patterns occur, each exactly once, with none missing and none duplicated. I verified this exhaustively against the sixteen-member set of four-digit strings over the alphabet {1,2}.

Column pattern (top to bottom)	Name	As binary (single=1)	Decimal
1111	Ogbè	1111	15
2222	Ọ̀yẹ̀kú	0000	0
2112	Ìwòrì	0110	6
1221	Òdí / Èdì	1001	9
1222	Ọ̀bàrà	1000	8
2221	Ọ̀kànràn	0001	1
1122	Ìrosùn	1100	12
2211	Òwónrín	0011	3
1112	Ògúndá	1110	14
2111	Ọ̀sá	0111	7
1121	Ìrẹtẹ̀	1101	13
1211	Òtúrá	1011	11
2212	Òtúrúpọ̀n	0010	2
2122	Ìká	0100	4
1212	Ọ̀ṣẹ́	1010	10
2121	Òfún	0101	5

The mark patterns in this table are Bascom's, recorded from his Ifẹ̀ informants and cross-checked against his corpus of 86 lists 1. The binary and decimal columns are conversions performed for this corpus under the mapping stated above, and are arithmetic rather than ethnographic.

Why 256, and why not some other number

The number is forced by three independent design choices, and changing any one changes the total.

Two states per position, because the physical procedure yields exactly two outcomes: one nut or two, concave or convex. This gives base 2.

Four positions per column, which is the choice that is not forced by anything mechanical. Nothing in the palm-nut procedure requires stopping at four. Four positions give 2^4 = 16 columns. Three would give 8, five would give 32. The system stops at four, and sixteen recurs throughout Ifá: sixteen palm nuts, sixteen principal Odù, sixteen verses per Odù as the minimum for initiation, sixteen cowries in the related system. Whether sixteen is prior to the four positions or a consequence of them is not something the sources settle.

Two columns, giving 16^2 = 256. Two columns is what turns a set of sixteen categories into a set of 256 ordered pairs, and it is what makes the naming compound.

The order matters, which is why the total is 256 and not 136. If Ogbè-on-right-with-Ọ̀yẹ̀kú-on-left were the same Odù as Ọ̀yẹ̀kú-on-right-with-Ogbè-on-left, the number of distinct figures would be the number of unordered pairs with repetition, which is (16 x 17)/2 = 136. Because Bascom documents that the two are different figures with different content 1, the count is the full 16 x 16 = 256. The system is explicitly non-commutative, and that non-commutativity roughly doubles its capacity.

Bascom's correction of an old overclaim

Older literature inflated the number badly, and Bascom addresses it directly because it was widespread enough to need refuting.

He notes that "many authors have claimed that the figures of Ifa number in the thousands," deriving from J. Johnson's account of sixteen principal odù, "256 or 4,096 odù altogether, and, 'according to some' 65,536." Farrow inflated it further by pointing to the many possible results of each of the eight drawings, which Bascom observes "is already accounted for in increasing the number from sixteen to 256." His conclusion: "As any reader may verify for himself, however, the system permits no more" 1

The numbers 4,096 and 65,536 are 16^3 and 16^4, that is, what you get by positing three or four columns instead of two. They are arithmetically coherent and ethnographically wrong. The system has two columns.

This matters beyond bookkeeping. The inflated figures still circulate, and they are sometimes offered as evidence of the system's sophistication. The accurate number is 256, and 256 is quite enough.

A structural property of the ranking

There is a genuine mathematical regularity in how the sixteen principal Odù are ordered, and it holds in both of the major competing orders documented in file 03. I checked it exhaustively rather than taking it from a source, so it is presented as arithmetic on Bascom's recorded patterns rather than as an ethnographic claim about intent.

Call a column pattern vertically symmetric if it reads the same top to bottom as bottom to top. Exactly four of the sixteen patterns have this property: 1111, 2222, 2112, 1221. Those four are precisely Ogbè, Ọ̀yẹ̀kú, Ìwòrì, and Òdí, which are precisely the first four Odù in the ranking, in both the Ifẹ̀ and the southwestern orders that Bascom tabulates 1.

The remaining twelve patterns form six mirror pairs, each pattern paired with its own top-to-bottom reversal. In both orders, every one of those six pairs occupies two adjacent positions in the ranking:

• Ọ̀bàrà 1222 with Ọ̀kànràn 2221
• Ìrosùn 1122 with Òwónrín 2211
• Ògúndá 1112 with Ọ̀sá 2111
• Ìrẹtẹ̀ 1121 with Òtúrá 1211
• Òtúrúpọ̀n 2212 with Ìká 2122
• Ọ̀ṣẹ́ 1212 with Òfún 2121

The two competing orders differ only in the sequence in which these pairs appear and in which member of a pair comes first. Their pairing structure is identical.

What this establishes is that the ordering is not arbitrary with respect to the figures' geometry: symmetric figures first, then mirror twins kept together. What it does not establish is intent. No source I found states that the ordering was constructed on this principle, and I am not asserting that it was. The regularity is real and checkable; its origin is not documented.

A worked example, from cast to signature

This traces the mechanics on an invented cast. It is an illustration of notation, not an instruction for practice, and it omits the invocations, the ritual context, and everything that makes an actual consultation what it is.

Suppose a diviner works the palm nuts. The right column is filled first, top to bottom, four valid trials:

1. Two nuts remain, so one mark. Position 1 of the right column: 1
2. Two nuts remain, so one mark. Position 2: 1
3. One nut remains, so two marks. Position 3: 2
4. One nut remains, so two marks. Position 4: 2

The right column reads 1122 top to bottom, which the table above names Ìrosùn.

Then the left column, again top to bottom:

5. Two nuts remain, one mark: 1
6. One nut remains, two marks: 2
7. Two nuts remain, one mark: 1
8. One nut remains, two marks: 2

The left column reads 1212, which is Ọ̀ṣẹ́.

The full signature is 1122|1212. Named right column first, this is Ìrosùn Ọ̀ṣẹ́. As eight bits under the mapping given above it is 11001010, decimal 202, and as an ordered pair of column values it is (12, 10).

Had the same eight marks appeared with the columns swapped, 1212|1122, the Odù would be Ọ̀ṣẹ́ Ìrosùn, a different figure with different verses. The eight physical marks are the same; the orientation is what distinguishes them.

The diviner would then recite the verses filed under Ìrosùn Ọ̀ṣẹ́ until the client stopped him at the relevant one.

What the binary structure does and does not license

This is the section to read carefully, because the structural facts above are frequently pushed past what they support.

Established. Each position is genuinely two-state. Four positions per column give sixteen columns, two columns give 256 Odù, and the count is exactly 256 rather than more. Order of columns is significant. The two casting methods are informationally equivalent. The mark patterns for the sixteen columns are complete and non-redundant. The system uses base 2 in the precise sense that its atomic unit has two states and its figures are strings over that unit. Bascom's own 2^n formulation is a scholar's description of the combinatorics, published in 1969 1.

Not established by anything above. That the Yorùbá performed arithmetic in base 2, which is a different claim from having a two-state notation. Nothing in the sources shows Odù being added, multiplied, or manipulated as numbers. The decimal values in the table are conversions this corpus performed, and there is no evidence any diviner treats Èjì Ogbè as fifteen.

That the ordering is a numerical sequence. It is not: read as binary the ranking runs 15, 0, 6, 9, 8, 1, 12, 3, 14, 7, 13, 11, 2, 4, 10, 5, which is not any recognisable numerical order. The ranking is by seniority in the tradition's own account, which is mythological and is discussed in file 03.

That single mark equals 1 and double equals 0 rather than the reverse. The tradition assigns no numeral to either mark. Both mappings appear in modern writing, and choosing one is a modern convention. This corpus chose one and labelled it as a choice.

That any of this makes Ifá a computer, or a proto-computer. It is an addressing scheme over a fixed corpus, with no operations defined on the addresses, no storage that is written to during use, and no computation performed on the values. Calling it a binary information system is defensible. Calling it computation is not, and file 07 takes the popular versions of this claim apart individually.

The structure is impressive on its own terms. A society without writing built a randomising address space of exactly 256 slots, with a physical procedure that reliably produces uniform two-state outputs, a rejection rule to discard invalid draws, a fixed orientation convention to make the address an ordered pair, and a naming scheme that lets any practitioner name any address from its marks. That is a real achievement in information design and it does not need inflating.

History and evolution

The binary structure of Ifá divination represents an oral information technology that predates European contact in West Africa. In its earliest documented forms preserved across oral traditions and archaeological records in Ilé-Ifẹ̀, the system established sixteen primary figures (ojú odù) and their 256 paired permutations 1. As the Ọ̀yọ́ Empire expanded across Yorùbáland during the seventeenth and eighteenth centuries, Ọ̀yọ́ standardized the regional administration of divination guilds and patronized babaláwo, diffusing uniform notation and ranking across expanding territorial networks 1.

The nineteenth-century Yorùbá civil wars and the collapse of the Ọ̀yọ́ capital shattered traditional state institutions, displacing lineages and dispersing diviners across southwestern Nigeria 1. Concurrently, Christian missionary contact in the mid-nineteenth century introduced aggressive campaigns against indigenous knowledge systems. Early missionary accounts, such as those by James Johnson, systematically misunderstood the binary notation, inflating the count of figures into thousands and framing the practice as heathen superstition rather than structured mnemonic inquiry 1.

Under British colonial rule in the late nineteenth and early twentieth centuries, colonial authorities marginalized Ifá under anti-witchcraft ordinances, though the internal mathematical structure and casting techniques remained preserved within hereditary secret societies 3. Following Nigerian independence in 1960, academic studies by scholars such as William Bascom and Wande Abimbola systematically documented the 256 Odù and corrected colonial mathematical distortions, demonstrating the rigorous binary logic of the eight-bit system [S1, S2].

Today, the 256 Odù system operates globally across West Africa and throughout the African diaspora, including Cuban Santería, Lucumí, and Brazilian Candomblé traditions 3. In 2008, UNESCO recognized the Ifá divination system as a Masterpiece of the Oral and Intangible Heritage of Humanity, acknowledging its continued vitality both in traditional community governance and in transnational digital networks 3.