The general approach underlying the EAGLES proposal follows the ET-7 (Heid &\ McNaught, 1991) proposal of looking for the basic phenomena at each level of linguistic description, going to a more granular level, and providing there a detailed set of features able to encode the relevant phenomena.
This approach is taken here over the set of European languages, trying to reach the same level of granularity for the description of each. For each language, the most common practices for lexicons and corpora were considered.
The consequence of the two approaches taken together (i.e. granularity and many languages) is that a large repository of potentially useful lexicon specifications is formed where all the features which are necessary for the description of the basic phenomena in the many languages considered are juxtaposed. Given the differences in the various language-specific systems, where each system has its own set of constraints, the large collections of features -- summarised in the tables at the beginning of the description of each grammatical category -- do not, and cannot, constitute a consistent system to be implemented as such, but are a redundant inventory of all the possible features relevant for that category across the different languages (this constitutes an ET-7 conformant large repository, according to the data pool model). Each language-specific system can afterwards be implemented as a specific application of the general redundant set, by picking up the features and values appropriate for its system.
At the general level of the large synoptic tables there is, by definition (i.e. by the very way in which they are constructed), no property of monotonicity (and no necessity for it), but, in contrast, there is redundancy and conflicting values may also be found.
It is only at the level of the language-specific instantiations -- considered as applications where the problem of the representation formalism also arises -- that monotonicity can be looked for. Applications of the general tables are given for most of the European languages as a reinforcement of the EAGLES proposal, and here -- in particular when these will become detailed for real lexicon building -- the different constraints have to be specified, as far as possible, between different features; the range of pertinent values for each feature have to be made explicit; and aspects of the hierarchical organisation of the features have to be solved (see the guidelines produced within the validation phase and the PAROLE contribution).
Within this approach, it is assumed that not all the values presented in the general tables are relevant for all the languages in the set under study. There will be cases in which some do not apply, and this has to be made clear in the language-specific applications.