In selecting the closest prior art, the first consideration is that it must be directed to the same purpose or effect as the invention. Otherwise, it cannot lead the skilled person in an obvious way to the claimed invention. According to T 606/89 the closest prior art for the purpose of objectively assessing inventive step was generally that which corresponded to a similar use requiring the minimum of structural and functional modifications (see T 574/88, T 606/89, T 686/91, T 834/91, T 482/92, T 298/93, T 59/96, T 730/96, T 650/01). In T 273/92 the board of appeal confirmed the established case law of the boards, according to which a document could not qualify as the closest prior art to an invention merely because of similarity in the composition of the products; its suitability for the desired use of the invention also had to be described (see also T 327/92). According to T 506/95, the closest prior art was therefore that most suitable for the purpose claimed by the invention, not that superficially showing structural similarities with the solution as claimed. Ideally, that purpose or objective should be something already mentioned in the prior art document as a goal worth achieving (T 298/93, T 859/03). The aim was that the assessment process should start from a situation as close as possible in reality to that encountered by the inventor. The real-world circumstances had to be taken into account. If it was not clear from this criterion what the closest prior art was, the problem and solution approach should be repeated taking possible alternative starting points (T 710/97, T 903/04).
In T 176/89 the board concluded that the closest prior art comprised two documents in combination with each other. It found that, exceptionally, the two documents had to be read in conjunction; they had the same patentee, largely the same inventors, and clearly related to the same set of tests. As a rule, however, when assessing inventive step, two documents should not be combined if, in the circumstances, their teaching is clearly contradictory (see also T 487/95).