Inelastic light scattering (i.e., Brillouin and Raman) can provide information about chemical bonding and coordination states, making it an efficient technique for studying amorphous matter in supercooled liquid (SCL) and melt regimes. Several studies have reported that, during crystallization (which occurs between these regimes), binary system glasses exhibit anomalous elastic behavior [i.e., simultaneous occurrence of increases in sound velocity (or elasticity), which is estimated by Brillouin spectra] and significant damping. Such behavior has also been reported in the low-energy vibration mode ( 1THz) of Raman scattering (known as the Boson peak, BP) during transition from SCL to crystals (SCL–crystal transition). Although the anomalies appear to be related to the crystallization phenomenon, in-depth analysis has not been conducted yet. Therefore, in this study, in order to obtain additional information concerning the anomaly in the crystallization regime, we prepared an appropriate sample for the in situ observation of the BP and examined the behavior of the BP under nonisothermal and isothermal heating conditions. For observing crystallization dynamics through inelastic light scattering, glass samples that show a strong tendency of homogeneous nucleation are preferable because the samples have a large interface, at which the SCL–crystal transition occurs. Since a BaSi2O5 glass possesses high nucleation ability and TiO2 is known to be an effective nucleation reagent in a silicate glass system, we prepared the glasses in which SiO2 was substituted by TiO2, i.e., BaTixSi2 xO5 (x 1⁄4 0{0:25), using a melt-quenching technique (melting condition: 1773K for 1 h). The as-quenched samples were annealed at each glass-transition temperature (Tg) [Fig. 1(b)] for 20 h to obtain nucleated samples. This was done because for glass with a homogeneous nucleation trend, the temperature at which the maximum nucleation rate is evident is close to Tg. 9) The samples were characterized by differential thermal analysis (DTA), transmission electron microscopy (TEM), and in situ inelastic light scattering observation. Detailed conditions for these analyses are described elsewhere. The spectra that were obtained in the low-energy region were reduced with respect to the Bose–Einstein factor and were fitted using the sum of the BP and the Gaussian function. This was done because the vibrational band, which results from the Ba–O bond, contributes to the spectra. The observed BP was analyzed using a log-normal function to evaluate the position of its maximum !BP and its full width at half maximum (FWHM). Figure 1 shows the results of thermal analysis in the studied samples. Here the samples with x 1⁄4 0:125 are representative of all the samples. In the as-quenched samples, their Tg increased with the value of x. The annealed samples at each Tg tended to show lower crystallization-peak temperatures (Tp) than the as-quenched samples [Figs. 1(a) and 1(b)]. In particular, the difference in Tp between the as-quenched and annealed samples (i.e., Tp) indicated the maximum value at x 1⁄4 0:125 (BaTi0:125Si1:875O5; Tp 1⁄4 73 K). This value was much larger than that of the base glass (BaSi2O5; Tp 1⁄4 40K). According to Marotta et al., oxide glass showing homogeneous nucleation and subjected to thermal treatment reveals a shift of Tp to a temperature that is lower than that of the nontreated glass because of the evolution of an excess of internal nuclei. These results suggest that the BaTi0:125Si1:875O5 glass possesses high nucleation ability. In addition, the annealed sample showed no structural development, and only the halo pattern was observed in the corresponding electron diffraction pattern [Fig. 1(c)]. We, therefore, selected the BaTi0:125Si1:875O5 glass as a test sample for the in situ observation. Figure 2 shows the result of the in situ measurements of the inelastic light scattering spectra of the test sample. We observed a BP around 1–4 THz, which is attributed to an excess of the vibrational density of states because of the nanometric heterogeneity/cohesive nanodomain. From the fitting, we estimated !BP to be 1:3THz at room temperature [Fig. 2(a)]. During heating (nonisothermal; 20K/min), the spectra showed no significant change up to 1273K. When the temperature was further increased, the BP revealed progressive broadening (or overdamping) in the isothermal stage (0–21min at 1273K; T=Tg 1:28). Eventually, a group of small peaks (closed circles) (as opposed to the BP) began to appear, indicating full crystallization of the test sample. In terms of the reduced temperature (T=Tg) dependence of !BP [Fig. 2(b)], !BP 0 0.1 0.2 970 980 99