When a carbonyl compound is treated with a strong base, the α position gets deprotonated to give a resonance-stabilized intermediate called an enolate. Enolates are ambident nucleophiles because they possess two nucleophilic sites that can attack an electrophile owing to the delocalization of the negative charge between the α carbon and oxygen atoms. When the oxygen atom attacks an electrophile, it is called O-attack, whereas electrophilic attack via the α carbon is known as C-attack.

C-attack is much more common than O-attack despite the negative charge being more localized on the oxygen atom. (Recall that for a set of contributing structures with a negative charge on different atoms, the contributing structure with the negative charge on the most electronegative atom will usually most closely resemble the true structure.) As a result, there are two distinct conventions in drawing the C-attack mechanism:

1. The oxyanion form, where the mechanism begins with the oxyanion contributing structure. This is technically a more accurate representation of the mechanism because the oxyanion form is closer to the true structure, but it involves an additional mechanism arrow to show the flow of electron density from oxygen to the α carbon.
2. The carbanion form, where the mechanism begins with the carbanion contributing structure. This is technically less accurate because the carbanion form is a less significant resonance contributor, but it is still a clear and valid representation so long as it is understood that the electron density of the delocalized system is actually concentrated on the oxygen atom.

The carbanion mechanism convention is more simplistic in representation because it needs fewer curved arrows, but both conventions are widely used.