Here are several cases of different arrangements of sound sources and how the hearing system handles finding where the sound comes from. Sometimes it works correctly and sometimes it gets confused.
The term "cogging" comes from an effect found in digital position encoder knobs. When the power is off, the knob turns freely. When the power is on, the knob stops and stays at each intended detent position. It acts like there is a spring detent on a cog on the shaft. The effect of a sound suddenly seeming to jump from one speaker to another is likened to the detent effect, so it is called cogging.
In monophonic play, there is only one speaker. The human hearing system correctly locates it at the location of the speaker. All of the sounds from the player come from this location.
The human hearing system notes the loudness and phase of the sound in each ear, and the time difference between the ears hearing the sounds. The system then makes analog computations to determine the direction and approximate range of the source. The system works for mono no matter where the sound source is.
In two-channel stereophonic play, there are two sources playing the same sound. Sound from each speaker travels to both ears of the listener. The human hearing system then tries to locate the actual source of the sound.
These combine to use the ratio of loudness of the two loudest sounds to locate the sound between the speakers according to the ratio of the loudnesses of the two sounds. This makes possible the panning of sounds between speakers by changing the relative loudness between them. The pan pot on a mixer makes the needed ratio changes.
The other sounds reaching the ear are taken by the hearing system to be echoes.
As in case 1, the ratio of the sounds from the left and right speakers determines the perceived location of the sound between them. But the phase shifting tells the listener the sound is behind him.
In this case, the hearing system takes the loudest sound reaching each ear as the actual sound source. These two loudest sounds are usually from the same speaker. The other sounds reaching the ears are taken to be echoes.
As long as that speaker produces the loudest version of that sound, the hearing system locates it as the sound source. When the other speaker becomes the loudest source, the hearing system suddenly locates that speaker as the sound source. The perceived location "cogs" from one speaker to the other, even though the levels were intended as a smooth pan from one speaker to the other.
If the listener turns his head to the left so the speakers are to the left and right of his face, then he will hear the smooth panning as it was intended to be.
In this case, the hearing system takes the loudest sound reaching each ear as the actual sound source. These two loudest sounds are usually from the same speaker. The other sounds reaching the ears are taken to be echoes.
As long as that speaker produces the loudest version of that sound, the hearing system locates it as the sound source. When the other speaker becomes the loudest source, the hearing system suddenly locates that speaker as the sound source. The perceived location "cogs" from one speaker to the other, even though the levels were intended as a smooth pan from one speaker to the other.
If the listener turns his head to the right so the speakers are to the left and right of his face, then he will hear the smooth panning as it was intended to be.
-----
Note that it the listener in Case 1 turns his head to he left, he will hear conditions as in Case 4.
Note that anyone using an unassisted soundbar who turns his head left will hear only the Case 4 conditions.
In this case, cogging goes on as in cases 3 and 4, but with more cog stops.
In this case, the speakers can combine to create a sound image as in cases 1 and 2.
The cogging occurs when all of the speakers making the major sound are on the same side (left or right) of the head.
It is often not noticed unless the sound is being panned.
If the sound is panned between three speakers on the same side of the head, it will cog twice, from speaker to speaker to speaker.
Cogging can occur with matrix systems. But it always occurs with discrete recordings.