The development of microscale and nanoscale devices has outpaced the development of metrology tools necessary for their complete characterization. In the area of thermal MEMS technology, accurate measurements across a broad range of temperatures with high spatial resolution are not trivial. Thermal MEMS are devices in which the control and manipulation of temperature is necessary to perform a desired function, and are used in actuation, chemical sensing, nanolithography, thermal data storage, biological reactions and power generation. In order to properly design for reliability and performance issues amongst these devices and verify modeling accuracy, the temperature distribution under device operating conditions must be experimentally determined. Raman spectroscopy provides absolute temperature measurements with spatial scales below 1 micron, which is sufficient for most MEMS devices. In this work, a detailed study of Raman spectroscopy as an optical thermal metrology tool was performed. It is shown that a calibration of the Stokes shift with temperature yields a linear calibration for measurements up to 1000?n polysilicon. These coefficients were determined for polysilicon processed under various conditions (575-620?B and P doping) to assess the effects of microstructural variations on Raman spectra. The Stokes peak was also shown to shift linearly with an applied pure bending stress. In order to make stress-independent thermometry measurements, the ratio of the Stokes to anti-Stokes signal intensities and the Stokes linewidth were calibrated over the same temperature range. Using the calibration data, Raman spectroscopy was implemented for the evaluation of temperature of thermal MEMS. Heated AFM cantilevers and micro-beam heaters were chosen due to their wide range of applications. Different thermal and mechanical boundary conditions were considered by studying both the beams and cantilevers, resulting in varying levels of thermal stress. By using the three calibrations in a complementary fashion, the validity of Raman thermometry was explored. Device temperatures of up to 650?nd their corresponding uncertainties were found, and used to verify FEA modeling. Effects of thermally induced stresses were taken into account and analyzed. Possible uncertainties such as laser heating, spatial and spectral resolution, light collection efficiency, measurement uncertainty, and instrumental drift were reported and elucidated.