The widely accepted mechanism of long chain branch (LCB) formation, i.e. random macromer incorporation, conflicts with many of the experimental trends observed in polyethylene production. This includes the response of LCB to common reactor variables, the surprisingly high LCB incorporation efficiency, and in particular the effect of metallocene structure on LCB levels. Many metallocenes are known to produce comparatively high amounts of LCB, while others produce little or none. In contrast to expectations from the widely accepted mechanism, this frequently has little to do with the site's ability to incorporate comonomer. Instead, LCB and SCB (short chain branch) incorporation are found to be relatively independent of each other. These problems are addressed and a modification of the LCB mechanism is proposed, that is intra- versus intermolecular macromer incorporation. The consequences of the proposed intramolecular insertion are discussed, including the effect of metallocene structure on LCB and SCB incorporation. The ability of a metallocene site to incorporate LCB may be mainly determined, not by its ability to incorporate comonomer, but rather by its ability to coordinate and hold on to terminated macromer. Consistent with intramolecular LCB formation, no evidence was found in slurry polymerization experiments of cross-insertion between sites, i.e., the incorporation by one site of macromer produced from another site. This has implications for the predicted architecture of polyethylene (e.g., Y's, trees, combs, etc.).