Determinations of phase boundaries and thermochemical parameters using conventional hydrothermal techniques of analysing the 'quench' products often result in large uncertainties in experimental data. This problem is especially troublesome when the phases involved undergo compositional variation during crystal growth, and possibly during experimental quenching, producing deviations in the mineral compositions from the actual bulk composition of the experimental starting material. Such compositional offsets can be difficult, if not impossible, to distinguish from the presence of immiscibility regions within the compositional field. As a consequence, it is important to focus attention on the process of nucleation and mineral growth to better understand the physiochemical behaviour of mineral species. A point of recent discussion has been the t rue c o m p o s i t i o n o f t r e m o l i t e (Ca2MgsSisO22(OH)2) grown in hydrothermal experiments. It has been virtually impossible to synthesize an on-composition tremolite experimentally (Jenkins, 1991; Pawley et al., 1993). This 'offcomposition' behaviour has been explained as a result of a narrow miscibility region in compositional space, while several crystallographic studies have assigned it to numerous 'ordered and disordered' variations in the 'b iopyribole ' type structure (Maresch et al., 1994). In this study we conducted experiments in a Hydrothermal Diamond Anvil Cell (HDAC) for in s i tu observations of the process of tremolite nucleation and growth. We used a modified version of the HDAC (Shen et al., 1992), which uses a pair of resistance heaters around the diamond anvils, capable of heating the cell to 1000~ We used a decarbonated mixture of chemical oxides (MgO, CaCO3, SiO2) in stoichiometric proportions for tremolite as the starting material for the watersaturated synthesis experiments. Additional silica (approx 6 wt.%) was added to ensure silica saturation in the system. The experiments were set above the talc stability and below the tremolite breakdown curve (Jenkins, 1987) at temperatures of 820 ~ • 20~ and pressure ranging between 2 6 kbar. A rhenium gasket was used as the sample chamber in which the oxide mixture and water was added leaving a visible air bubble. The bubble was used to determine the Th (homogenization temperature) of water, which was then used to estimate pressure at a given temperature following the corresponding water isochore. Initial pressure estimates were made using this technique during heating as well as quenching of the experiment. Differences in values between these determinations were primarily due to volume changes of the sample chamber. The crystallizing phases were monitored in-situ, visually as well as using Raman spectroscopy. The run products were first characterized using Raman spectra referenced to a natural Barrie Tremolite spectrum and later examined in a SEM to record the texture and estimate the synthesis yield. These 'time-series' show rapid crystallization in the form of radiating fibers surrounding the starting oxide mixture. With increasing duration of the experiment, the crystals were observed to grow from a fibrous form to an acicular form. In-situ laser Raman spectra obtained for the synthesized material indicated spectral peaks characterstic of amphibole at 400 cm -1 and 650 cm -1. For short duration experiments, these peaks appeared as doublets, which closely correspond to the vibrational spectral peaks of a pyroxene (Fig. 1). For successively longer duration experiments, the 'twin-peak' pattern disappeared to produce a spectral pattern similar to the reference tremolite. These observations provide the physical evidence for a compositional (and structural) transformation of tremolite during nucleation and growth. Our results show that the reason for the 'off composition' behaviour of experimentally produced tremolite is the formation