Tone plays a crucial role in distinguishing lexical meanings in Mandarin. In the 3rd tone sandhi, whether and to what extent does the tonal shift from the T3 to the T2 causes semantic confusion? Previous studies from acoustic and perceptual perspectives have not yielded a definitive conclusion. This paper attempts to investigate the information mechanism behind the phenomenon of the 3rd tone sandhi from the perspective of the linguistic communicative function. We examine the number of lexical minimal pairs, lexical similarity, and sentence information loss after sandhi at both lexical and sentence levels. The results reveal that at the lexical level, the number of minimal pairs formed between T3T3 and T2T3 in dictionaries of different sizes is the smallest; at the sentence level, sentences with no information loss are most numerous when T3 is confused with T2, and both the average and ranking of information loss are the lowest. These indicate that the communicative risk caused by confusing T3 and T2 was small. These findings provide a new perspective on the mechanism of the 3rd tone sandhi.