Because of their unique properties, shape memory alloys have been used in many ®elds to replace traditional metals. Although some constitutive models have been developed to describe the thermomechanical behavior of shape memory alloys, the most practical approach for engineering applications is still phenomenological models. Among them, the model initiated by Tanaka [1], developed by Liang and Rogers [2] and further modi®ed by Brinson [3] is probably the most popular one used by engineers. In this model, phase transformation temperatures (austenite ®nish temperature A f , austenite start temperature As, martensite start temperature Ms and martensite ®nish temperature M f ) are assumed to be linear functions of external stress, i.e., the higher the external stress, the higher the transformation temperatures. Another assumption of this model is that the transformation process follows a certain function (exponential or cosine). However, such treatment does not take the effects of internal stress and martensite variants into consideration. As pointed out by many researchers (for example, [4] and [5]), internal stress indeed plays a very important role in both forward and reverse transformations. Although in Brinson's model [3], two separated fractions are used to distinguish twinned and detwinned martensite, the effect of martensite variants still has not been given enough attention. We exempli®ed the effects of internal stress and martensite variants on the transformation process and As=A f by means of the differential scanning calorimeter (DSC) test. The shape memory alloy used was 1 mm diameter NiTi wire, which was supplied by Thomas Bolton Limited. Details of the preparation are given in [6]. Two kinds of samples were tested. One was cut from a wire that had been stretched by loading to 350 N (sample 1) at a strain rate of 2:6 3 10y4 and an ambient temperature of about 20 8C, while the other was cut from a wire without any pre-stretch (sample 2). To determine if there was any uncoverable deformation during loading=unloading, we heated pre-stretched wire to 150 8C (the DSC test shows that this NiTi alloy is austenite at this temperature). No noticeable unrecoverable deformation was observed. The loading=unloading test was then repeated. As shown in Fig. 1, strain versus force relations of the two tests are almost the same. It suggests that there is almost only elastic deformation besides phase transformation. DSC tests were carried out at 10 8C=min. Both samples were heated from 20 to 150 8C. As shown in Fig. 2, temperature versus heat ow relations of samples 1 and 2 are very different.