There are several kinds of language that using in newspaper. Its depend on the character of informations. According to Hoed (1976a), the various of languages in newspaper base on the ideas of sociolinguistics field that regarded language in reality is not monolitic, but consist of several variations. One of the inner problem in Indonesian mass media and become the focus of this research is the cohession. The goals of this research, which the tittle is "Kohesi dalam Bahasa Indonesia: Kajian Atas Teks Tajuk Rencana Lima Surat Kabar Harlan di Jakarta", is to describe the editorial cohession (tajuk rencana) in Indonesian daily newspapers. Through this research, I hope that I am be able to describe (i) cohession mecanism as the tie of propositions in editorial discourse; (ii) language units which's used as cohession marker in editorial text; and (iii) the kinds of cohession marker which's frequently used in editorial text and their causal factors. This research use Halliday and Hasan theory (1979) which combined with Cruse theory (1986). In this research I found that the same cohession is not always using in editorial text eventhough in the same newspaper. But, it's an unabsolutely using the all cohesion marker in editorial. The result of the research show that there is a tendency in using cohession marker which depend on the "topic" of the news. The language units that places as cohession marker in editorial are ia, dia, mereka, -nya, itu, ini, ini, itu, begini, and di sini; conjunction sejak, karenanya, tetapi, akan tetapi, namun, semen tare itu, clan, ketimbang, bahwa, sebab, sehingga, meskipun; and OIeh karena itu. And, I also found the using of noun, noun phrase, verb, verb phrase, and clause in the editorial. Through the comprehensive analysis of the six topic of editorial, which represented by one every topic, this research found 254 cohession marker, both gramatical or lexical. The lexical cohession marker repetition is the most frequently use, i.e. 117 (45.9%). Then the reference 53 {21.1%), collocation 28 (11%), conjunctive relation 20 (7.8%), substitution 8 (3.1%), ellipsis 8 (3.1%), paronymy 7 (2.8%), synonymy 6 (2.4%), hyponymy 5 (2%), and meronymy 2 (0.8%).