: Metal-oxide gas sensors are widely used in gas detection applications due to their low cost and high sensitivity. These sensors also respond to a wide range of chemical compounds. However, lack of selectivity and their operation at elevated temperatures are the main limiting factors preventing their use in portable gas analysis instruments. Temperature modulation is an effective method to overcome these drawbacks. In this method the operating temperature of the sensor is not constant; instead it is varied according to a temporal pattern. Here, the transient responses of a generic metal-oxide are analyzed to find an optimum pattern for temperature modulation. Gas recognition time, power consumption and discriminability are the main three criteria used for pattern optimization. First, the results of experiments on thermal measurements were used to model the thermal behavior of the sensing pallet. Using the inverse model, the appropriate heater voltage patterns required for generating the desired temperature patterns were derived. The three basic patterns of interest are ramp, exponential and step functions. The sensor was thermally excited using the basic temperature patterns. Experiments were repeated with different combination of characteristic parameters, e.g. amplitude, time constant and slope. The discriminability of the resulted transient responses were compared based on their spatial variation in the PCA feature space. Those resulted from step modulation were the best for optimized gas identification. It was also shown that normalized transient responses can be successfully employed for quantitative analysis of the target gases. The results of basic patterns were also used as a reference to generate combinational temperature modulation patterns. Keywords: metal oxide gas sensors, temperature modulation, thermal transient 1 response, gas analysis.