In certain cases when dealing with more phylosophic issues, than researches of >20 yrs ago may necessary, and real library available at universities. However, current references are more reliable for students pursuing their diploma or producing new researches .

I've found that Google Scholar does a good job of indexing even older works...the OCR they employ is quite effective at digitizing older journals. That said, the results do skew toward the 70s and onward...one must be diligent to find all the seminal papers in a field!

Recent literature is the outcome of old literature. So i think, going back to more than 20 yrs for literature review is not worthable.

In my experience, Google Scholar is excellent, Web of Science is good, but many others are woeful at indexing once you go back to the 1990's. Scopus is especially useless, I note that it halves my h-index by ignoring earlier work.

Thanks to all of you for the useful responses. First of all, let me say that, as my carrer started 35 years ago, I see quite well the problem underlined by Paul. The older papers somehow disappear from the memory. In order to keep some of them alive I (and many others also) wrote the books. But the books may synthesize only a few aspects of a paper. Second problem, shown by Volker and Imran is whether the present science is a continuation of older one? I do not think so. In my field some methods somehow became popular and some others disapeared entirely. The reasons are not always logical or even reasonable. Finally, I found myself quite often reading a paper which repeats exactly something what has been done 25 years ago and the authors do not cite the source because they did read it.

The problem is there, I am sure, the more so since there definitely is some recurring nature to scientific trends - metal foams is an example I am familiar with in this respect. If every publication has a backlog of about 20 years (the time from which it draws the references it cites), it may help to "step back" via the "thread of references" offered thus: If you know which paper you are looking for, you will most likely find it in some digital repository. However, this path is subject to some filtering effects, too.

Thus it seems we don't really know yet how to avoid things being invited several times.

I think you are wrong. The evolution of technology is one thing and scientific ideas being behind the scientific papers is something else.

Lech, I think the answers you will be getting will very much depend on what field/area of science you are talking about. Physics? Chemistry? Biochemistry? Applied material sciences? General biology>? Molecular biology? You get the idea... It will even more depend on WHAT it is that you are looking for and what the purpose of the search is.

In biological and medical sciences, if I want to get a quick overview of what has been going on in a certain area of interest, many of the cited databases do a good job. I personally think PubMed is excellent and quite comprehensive for life sciences. For more recent articles I also like "Scirus". For chemistry things get trickier. A quick search for a topic of interest can be done on Pubchem, Chemspider, etc. but if the search for a particular molecule or substance does not turn up anything, most chemists would know to turn to CAS or even Beilstein, which goes back all the way to 19th century literature.

Also, if I were to write a review on a particular topic I sure would spend more effort on researching the literature than if I just wanted to find out what's new. Most public electronic databases are really meant for the latter purpose and don't claim to be comprehensive.

As for you mentioning "reading a paper which repeats exactly something what has been done 25 years ago and the authors do not cite the source because they did read it", there is a word for that: plagiarism! Not having read prior literature or having failed to do a literature search is not an excuse. While, unfortunately, it happens, the paper should be pulled or corrected.

Frank, thank you for your opinion. I used to work for some time with the chemists and in chemical institutes and I agree with you that there are a couple of databases covering correctly the field back to XIX century. In the field of coatings technology in which I am active, we have a strong physical part mainly of fluid mechanics, heat transfer as well as science of materials.

I use ScienceDirect and Scopus when I do the literature research and sometimes I miss the older papers. Sometimes, I can find them in the bibliography of the available electronically papers.

Finally the citation "reading a paper which repeats exactly something what has been done 25 years ago and the authors do not cite the source because they did NOT read it" missed NOT. I was worrying about repeating the studies and not about plagiarism.