The difficulties faced by non-native writers in producing accurate, effective 

expository texts in English have prompted a multitude of studies on the elements 

that constitute well-written academic prose, and the ideal way to teach them to 

students learning English for use in academic contexts. A significant number of 

these investigations harness the power of computers to analyze language corpora 

— large collections of digitally stored, naturally occurring texts — with the aim of 

establishing linguistic and textual patterns and developing systematic descriptions 

of these patterns. 

         One of the most important findings revealed by corpus-based language stud

ies is the fact that, instead of constantly making new combinations of individual 

words, native speakers often depend on a stock of prefabricated, semi-automatic 

word chunks (Sinclair 1991). These results have led researchers to look beyond the 

single word in language description and pay attention to units of meaning com

prising multiple words.