Think about researchers who reviewed hundreds of papers that they had to access manually! UT Austin, Texas in 1987 had around 6 millions items distributed in several departmental libraries. Some of these were several storyed library buildings. There was also a huge 6 floor central library. For example, In Engineering building library each engineering discipline had their own encyclopedia type abstracts volumes. Computer and control abstracts had around 50 volumes (encyclopedia like thick volumes containing the abstracts of all papers published). Volume 1 and 2 contained the index of citations only. From the alphabetical index search you identified the particular volume, say volume #34 containing abstracts of a particular area. Then you pulled out volume#34 to read the abstracts and titles. If you thought the abstract was any good or relevant and the article was worth reading, then you noted the reference in a diary. Once you have noted these 20-30 worth reading citations, the next stage was to get the article. Fortunately, library catalog was computerized and UT Austin had a huge subscription of journals. Online catalog told you whether whether the journal was available and at which library, and which floor and which shelf. Generally from 20 citations you found 5-7 not available on campus. For these you needed to go to inter University loan service that typically took from 2weeks to two months to get the photo copy from a partner University. If the journal was available on campus, you went to that physical library location to find the volume of journal. You were lucky if the journal was available on shelf. Often another student may have taken out the volume and placed it in the library somewhere. For that you had to contact the library staff to search the item and hold it for you which may take a few days. Finally, you find the relevant volume and read the article. If it is really good and relevant you got to get the photocopy from the machines where you had to insert a prepaid cartridge to get the photocopies.