The theoretical framework of this thesis is based on the seminal work of Halliday and Hasan (1976). The main goal of the thesis is to reveal what types of grammatical cohesive relations are displayed in argumentative essays of academic written English. To achieve this purpose the International Corpus of Learner English (ICLE) has been chosen to study how Norwegian and Russian learners of English construct their compositions. Four major types of grammatical cohesion have been studied: reference, substitution, ellipsis and conjunction. Argumentative essays by Norwegian and Russian learners have been analysed in order to show how various grammatical elements function as cohesive links for sentences and independent clauses. The examination of grammatical cohesion shows that argumentative essays do not differ greatly in the number of cohesive items. A difference is, however, observed in the way these items signal different types of cohesion. Both groups of argumentative essays display a range of cohesive ties that link sentences and independent clauses. However the ties are not evenly distributed. The evidence of the examination suggests that reference and conjunction are the most common types of grammatical cohesion, whereas substitution and ellipsis are not represented widely. Both learner groups use extensively three exponents of anaphoric reference, namely personal, possessive and demonstrative. The examination of demonstrative reference shows that determiners work together with lexical cohesion. An overall impression is that lexical cohesion dominates in the essays of Russian learners. The findings support the hypothesis about the use of the. Both learner groups use the article the for presuming definite meaning. Norwegian learners do not confuse the use of the definite article. The essays of Russian learners display a different picture. The definite article is used to establish cohesive relations between sentences. However, this type of grammatical cohesion is not fully represented. Some learners make the wrong choice of English articles and such instances are not taken into account. The findings support the hypothesis about the use of and. Norwegian learners favour the use of and. It occurs initially to link two adjacent sentences or independent clauses. The same conjunction does not often appear initially in the essays by Russian learners. This position is taken by besides. It is possibly because of cultural influence (or language teaching), since and is not considered to be correctly used at sentence boundaries. The most surprising finding is that neither of the learner groups organize information in terms of temporal relations. Norwegian learners introduce a few conjunctions that signal the initial stage of a writer’s arguments or a concluding remark. Russian learners are reluctant to organize their essays in a sequence of arguments. They neither summarize nor give a resume what has been said. Instead they highlight the end of an argument by an expression of personal opinion. It is assumed that lexical cohesion is more widespread than the use of grammatical reference in the essays by Russian learners. It seems significant that grammatical cohesion should be studied in comparison with lexical cohesive items in a future study of cohesion in learner texts. Furthermore, contrastive studies and analyses of argumentative essays written by the native speakers of English would expand the framework and they would reveal more about the nature of grammatical cohesion as well as of the degree of success of the advanced learners of English.