Although a substantial body of research has supported that social identity and in-group identity play an important role in group-reference effect (GRE) on memory, there is still a need to continue the exploration of processing characteristics of memory with reference to some other social identities (Johnson et al., 2002). College students’ multiple social identities, for example, schools or nationalities they belong to, were usually investigated in previous studies, however, their major identity, an effective and vital group classification criteria, was often overlooked. Besides, typically researchers have looked at SRE (self-reference effect) and GRE in a relatively isolated way, which possibly goes against discovery of relationship between the self of individual and group levels (Bennett et al., 2010). Moreover, in-group identity has been proved to be an important influential factor of GRE. Through two experiments, combining paradigms of SR (self-reference) with GR (group-reference), the aim of present study was to examine whether there was a self- or major-reference effect (MRE) in college students’ memory and to compare the strength of them, then to explore the role of major identity and to contrast the differences of personal self and collective self. In experiment 1, there were three reference conditions (self, major, semantic) with 28 Chinese undergraduates who majored in psychology as participants. All of them had no experience of minor. Each condition was assigned 20 trait words with a total of 60. In study phase, participants were asked to determine to what extent the current word was suited to describe the referential object. For instance, “To what extent the word is suited to describe myself?” or “To what extent the word is suited to describe people who major in psychology?” and so on. Participants needed to complete a 5-point Likert assessment by pressing the keyboard. In recognition phase, 120 words were presented one by one and half of them had appeared in study phase. The participants’ task was to judge whether each word had been presented before. If it had, then they were instructed to make a ‘‘remember” or “know” response (R/K response) according to their retrieve levels. In experiment 2, participants who majored in economics and an out-group condition were added. Accordingly, trait words in both study and recognition phases increased to a total number of 160, and 64 Chinese undergraduates who majored in psychology and economics participated in this experiment. There were four reference conditions, namely self, in-group, out-group and semantic. Ultimately, a questionnaire was used to measure the level of participants’ major identity. The procedure of experiment 2 was akin to experiment 1. Total recognition rate, recognition rate on R response and discrimination index d＇were acted as dependent variables. A one-way repeated measures analysis of variance (ANOVA) was conducted on these three dependent variables. The results showed that, when encoded with reference to self, the words in both experiments were better remembered than any other condition. The performances in GR encoding tasks were also better than out-group condition (experiment 1) and semantic conditions (experiment 1 and 2). Besides, major identity affected the memory performance with reference to in-group, rather than self, out-group, and semantic conditions. When the score of major identity was high, a better score on memory of in-group referential words was obtained than out-group ones. However, when it was low, there was no difference between them. These results demonstrate that as a result of elaborative and organizational processing (Klein, 2012b), there is a robust SRE with respect to Chinese college students’ memory, as well as a MRE. In contrast, SRE is stronger than MRE. Furthermore, the level of college students’ major identity might play an important role in MRE.