These videos demonstrate how the molecular mechanism of recombination can also be responsible for gene conversion. The organism in which gene conversion has been best studied is a mold that keeps its spores in a sac called an "ascus". As meiosis progresses, the cells are kept in a linear array which allows the scientist to follow the fate of each cell. You can map a gene with respect to its position from a centromere using this system.
But gene conversion, not mapping, is what I want to describe. When meiosis occurs, you should have equal numbers of each kind of allele when you end. For example, if you have a B and a b allele of a gene, the end product of meiosis, four cells, will contain equal numbers of B and b two cells will be B and two will be b. Recombination will assort the alleles, but you're not gaining or losing any genetic material. If the haploid cells at the end of meiosis undergo mitosis, you'll double the number of both alleles: you'll get four cells that contain B and four that contain b.
Robin Holliday, a geneticist, was studying asci and mapping out how the alleles segregated. He noticed a strange phenomenon, though... sometimes instead of getting 4 B alleles and 4 b alleles, he got 5 B and 3 b. Sometimes it was 6 B and 2 b. Even the reverse happened: 3 B and 5 b. Sometimes it was 2 B and 6 b. Apparently one of the alleles got changed - converted - to the other allele in the cross.
Holliday recognized that the presence of both alleles together in the zygote might lead to them interacting. Crossover is when nonsister chromatids are more likely to interact, and he thought of a model where the double-helix of each chromatid might unzip and then base-pair with its nonsister partner. They'd have essentially the same nucleotide order, except for the region that differs to make them allelic and not identical, and so base-pairing can connect these nonsisters together. They'll be cut apart to separate, and depending how you cut them, you might cause a crossover event in which new allele combinations on a chromosome result.
Here's a video (courtesy of Brooker's Genetics, 2nd edition, McGraw-Hill)
... also ...
Holliday's model has inherent symmetry, though, and the converted asci don't typically have symmetry in their numbers. 5:3 is not symmetrical. Matthew Meselson and his colleague Charles Radding modified Holliday's theory by introducing assymetry: only one strand is cut and it dislodges the partner strand of the nonsister chromatid.
A further model, double-strand break repair, is a further refinement of the crossover model and is considered to be the most likely mechanism, although from what I hear, there are differences between species.
I like this concept as it demonstrates the progression of science, and how we can use concepts from different parts of the discipline to inform the other parts. This is a great combination of how classical and molecular genetics can cross-pollinate!