We examine the borate group intermixing in a series of 25 borosilicate (BS) glasses from the [0.5M (2) O−0.5Na 2 O]−B 2 O 3 −SiO 2 systems with M = {Li, K, Rb, Mg, Ca} along with ternary K 2 O−B 2 O 3 −SiO 2 and Na 2 O− B 2 O 3 −SiO 2 glasses by double-quantum−single-quantum (2Q−1Q) 11 B correlation nuclear magnetic resonance (NMR) experiments. The alterations of the fractional populations of B [3] −O−B [3] , B [3] −O−B [4] , and B [4] −O−B [4] linkages in the glass networks were monitored for variable n Si /n B molar ratios, nonbridging O contents of the glass, and the (average) cation field strength (CFS) of the M z+ /Na + network modifiers. A significant B [4] −O−B [4] bonding is observed in all glasses, thereby conclusively demonstrating that the normally assumed "BO 4 −BO 4 avoidance" is far from strict in BS glasses, regardless of the M z+ field strength. We show that the degree of B [4] −O−B [4] bonding depends foremost on its underlying BO 4 population and to a lesser degree on the NBO content of the glass; we also provide a straightforward prediction of the B [4] −O−B [4] population in an arbitrary BS glass from parameters readily obtained by routine 11 B NMR. The propensity for forming B [4] −O−B [4] linkages increases concurrently with either the CFS or the amount of glass network modifiers, roughly scaling as the square root of the "effective CFS" that encompasses both parameters. Although BO 3 −BO 3 and BO 3 −BO 4 pairs remain favored throughout all examined BS glass networks, the borate group intermixing randomizes significantly for increasing effective CFS, out of which the amount and charge of the glass-network modifier cations dominate over their size. Our results are discussed in relation to the two prevailing but formally mutually exclusive "random network" and "superstructural unit" models of borate and BS glasses.