In this thesis I explore the complexity of FL (Foreign Language) reading through qualitatively and quantitatively analysing the forms, ways, and mechanisms applied by adult readers at tertiary university education level to construct meaning in an ESP/EAP (English for Specific and Academic Purposes) multilingual educational context at the Eduardo Mondlane University (UEM), in Mozambique.I attempt to answer to the following research questions (a) What do learners and users of English in an EAP context resort to construct meaning from text?, (b) Are reading strategies used effectively by these learners to attain comprehension? and (c) Are these learners aware of their use of reading strategies? I identify reading strategies using a holistic eclectic research methodology that includes a Needs Analysis, a Reading Comprehension Test (IELTS), Questionnaires and a Think aloud Method (TAM). This holistic approach anchors partially on Bernhardt’s 2005 compensatory model of reading in a second language, which calls for answers to a ‘50% unexplained variance’, some of which I would like to resolve with this study.My study revealed that the participants are taught reading strategies formally but with the use of old outdated textbooks in the undergraduate courses at UEM, which showed to be characterized by a restrictedness to word and sentence level analysis (West, 1998), (ii) a descriptive yet not explanatory nature (Robinson, 1991), and (iii) and a pattern in which a long non-authentic specialist reading passage begins most lessons/units followed by exercises (Dudley-Evans & St. John, 1998). Further, I have revealed a lack of collaboration among various concerned stakeholders at UEM, including students, subject teachers, institutional administrators and EAP teachers to find answers to the traits revolving around course design and its improvement (Tajinoa, James, & Kijimac, 2005). I also discovered that language practitioners devalued the central idea posited by a Needs Analysis, and this critical significant incident has helped us to see things in a new way and thus develop our understanding (Kerfoot and Winberg, 1997). All of this clearly calls for a structured analysis of the entire system at UEM.The reading comprehension test (RCT) revealed a gloomy picture where participants failed to construe meaning adequately, especially with respect to higher order reading skills. Nevertheless the questionnaires and the think aloud results showed evident awareness of reading strategies involving a chief use of metacognitive strategies, and a high frequency use of cognitive and supply strategies.I therefore believe that I have holistically and synergistically brought to light some explanations and suggestions of certain variables that could be used to fill in the gaps of Bernhardt’s (2005, 2011) three dimensional model and as such part of the ‘50% unexplained variance’. I have concluded that in the population studied there is both self-reported and evident use of a battery of reading strategies, given that all participants (weak and strong) used (almost) all reading strategies, be it in different frequencies than L1 readers, and also used those known to be unique for biliterate and multilingual FL readers, i.e. code switching, translation and the use of cognates, and a novel supply strategy, i.e. sight-translation. Nevertheless, there does not seem to be an apparent strong relationship between these strategies and text comprehension and task performance. Clearly when for the participants trust in the target language failed, their most familiar reading language (Portuguese) was used to resolve conflicting information, to predict and to confirm meaning, or to question oneself. This suggested a sign of lack of an adequate L2 language threshold or, perhaps the ‘accustomed’ use of a familiar means, Portuguese, as the communication and comprehension vehicle. In conclusion, their poor reading comprehension test results could be blamed upon a lack of L2 linguistic knowledge, of L2 higher order comprehension skills and of knowledge of text structure that triggers a strategy use rather different from L1 readers.