This study uses Raman, Si-29 magic-angle spinning nuclear magnetic resonance (MAS NMR) and O-17 triple quantum (3Q) MAS NMR spectroscopies on K2Si4O9 glasses quenched from high pressure melts (5.7 and 8 GPa) and on high-pressure crystalline phases of K2Si4O9-wadeite and CaSi2O5 to investigate the structural mechanisms that create high coordinated silicon. The effect of decompression on the glass structure was also investigated by varying the decompression rates after temperature quench. The spectrum of crystalline triclinic CaSi2O5 phase clearly demonstrates that the Si-[4]-O-Si-[5] species do not show a distinct signature in O-17 3QMAS NMR and their signal can either be represented as "Si-[4]-O-Si-[4]_like" or "Si-[14]-O-Si-[6]-like" species, depending on the local environment of the oxygen. This suggests that Si-coordination should be directly investigated by Si-29 NMR and not inferred from O-17 NMR spectra. Additionally, based on the comparison of percentages of structural species measured with O-17 NMR to those expected from the Si-29 spectra, it seems that most, if not all, Si-[4]-O-Si-[5] in rapidly decompressed K2Si4O9 glasses are represented as Si-[4]-O-Si-[4], not Si-[4]-O-Si-[6]. These results were used to successfully test previously proposed mechanisms for the generation of high-coordinated Si (Q(3)+Q(4)=Si-[5]+Q(4) and 2Q(3)+Q(4)=Si-[6]+2Q(4)) when NBO are present. Spectra from glasses that were decompressed more slowly (conventionally) show that there are small structural differences between the glasses with different rates of decompression. Based on Si-29 and O-17 data, a small percentage of Si-[6] converts to Si-[5] (similar to1% of the total Si) by the reversal of the previously mentioned mechanism. Interestingly, this small structural change shows a relatively large effect on the Raman spectra, which suggests that the effect of decompression on silicate glass structure may need re-evaluation with additional in-situ studies. (C) 2004 Elsevier B.V. All rights reserved.