Fortunately, the investigators of binary stars were generally very attentive to multiplicity higher than 2 and provided corresponding comments. These comments were amply used in the selection of objects for MSC. In particular, we scanned the comments to the Washington double star catalogue (Worley & Douglas 1984, hereafter WDS), to the Fourth catalogue of orbital elements of visual binary stars (Worley & Heintz 1983, hereafter VBO) and the Eighth catalogue of spectroscopic orbits (Batten et al. 1989, hereafter SB). These two latter catalogues were also intercompared directly to find objects in common.
The Third preliminary version of the Catalogue of nearby stars (Gliese & Jahreiss 1991) was scanned to find multiple systems and this selection was checked later with the catalogue of Poveda et al. (1994) based on the same source. Notes to the Fifth edition of the Yale Bright star catalogue (Hoffleit & Warren 1991) were used as well for initial selection of objects. Multiple stars mentioned in the Second catalogue of speckle interferometric measurements of binary stars (McAlister & Hartkopf 1988) were added.
Naturally, the multiple systems described in the current literature or studied by the author were also included. In 1994 a program of radial velocity measurements of late-type components of multiple systems with correlation radial velocity spectrometer was started, and its results (tests of physical relation between visual components and discoveries of new close sub-systems) are incorporated into MSC.
Finally, the multiple star lists mentioned in the Introduction were merged into MSC. From the catalogue of Popovic (1991) we took only systems which he considered to be physical. Popovic used the WDS as input source, but this latter in fact contains many more visual multiples. Dommanget (1988) gives the number of visual multiples in CCDM (which contains approximately the same objects as WDS) as 5800. The physical relation between the components of these systems is rarely proved, so it seemed reasonable to restrict the present study to the list of Popovic until the nature of the remaining systems is elucidated, e.g. by the HIPPARCOS observations.
It can be stated that MSC is "complete'' in the sense of taking into account most of available sources: any known multiple system may be found in MSC with a good probability, with exception of visual multiples as explained above.
The available data on multiple stars are very inhomogeneous in nature and quality, and, consequently, no uniform and coherent description of all known multiple systems is possible. Our approach is to give a reasonably complete set of parameters for each system, bearing in mind that the reliability and quality of these data is very different, ranging from crude and uncertain estimates to good measurements. Answering the potential criticism, we feel that it is actually the only possible way to combine the information, and a necessary step towards the definition of future observing programs that would provide data of better quality and homogeneity.
Figure 1:
Mobile diagram of the quintuple system 
Ori and the
corresponding level codes. Visual components are designated by
upper-case
letters, spectroscopic components by second small-case letters.
Separations are given for visual systems, periods for spectroscopic
systems
Our aim is a catalogue of physical systems containing at least 3 stars. In case of wide visual double stars it must be proven that their components are not optical. The criteria used to identify optical components are detailed in Sect. 3.1. Wide systems with uncertain status were not included in our final list, but were retained in the database for further observations. It turned out that a large fraction of systems listed as physical by Popovic (1991) contains optical companions.
As for the close spectroscopic sub-systems, their reality is not always unquestionable, too. In the past radial velocity observations of poor quality or with different systematic offsets has led to the identification of spectroscopic binaries which are in fact single. A good example is Gl 4AB, both components of which are noted as spectroscopic binaries in the old catalogue of nearby stars (Gliese 1969) but have been found to have constant radial velocity (Tokovinin 1992). So, the stars without computed spectroscopic orbits are considered as spectroscopic binaries only when the velocity variability is confirmed by modern observations.
We caution the potential users of MSC that a few spurious multiple systems may still be listed despite the efforts to delete them.
Several stars with common distance and common spatial motion
do not necessarily form a multiple stellar system. They may be
just members of a stellar cluster, association or moving
cluster. Is there a formal distinction between multiple
stars and clusters? The only significant difference concerns
the type of motions, quasi-keplerian for hierarchical
multiple stars and stochastic for other stellar groups.
This criterion is however both questionable and impractical. The
possibility of quasi-periodic motions is easier to check.
It is known that wide pairs with separations
greater than 
A.U. are disrupted by encounters with
stars and molecular clouds (e.g. Weinberg & Wasserman 1988), i.e. are
not stable binaries. This consideration lead Abt (1988) to propose a
distinction between wide binary stars and co-moving pairs. It
seems thus inevitable to impose an upper limit on the separations of
multiple star components in order to distinguish them from other
stellar groups. The statistical difference between multiple stars
and open clusters was discussed by Dommanget (1977).
An interesting conclusion was reached by Larson (1995) from a study
of the clustering properties of young pre-main sequence stars. He
found that for separations r greater than 0.04 pc the probability of
finding a companion is represented by a power law which is related to
the fractal structure of star-forming clouds. For smaller separations
this probability is proportional to 
as typical for multiple
stars. This finding strengthens the idea of limiting component
separation that distinguishes multiple stars from clusters and
associations.
The "Larson limit'' is of the same order as the stability limit and
corresponds to an orbital period of 
years for 1 solar
mass system. This period was used in MSC compilation as a guideline
for rejection of wide pairs. However, no strong limit was
imposed and some systems with longer periods were still included.
It must be kept in mind that the separation between the members of the
Trapezium system is less than the limit
quoted above, and still this system is just the dense core of a
stellar cluster. Until the relation between multiple systems and
clusters becomes clearer, it is safe to adopt a somewhat relaxed
long period limit and to keep in MSC non-hierarchical
systems.
The majority of multiple systems are hierarchical and so can be considered as a combination of binary sub-systems. Evans (1968) proposed to describe hierarchical multiple systems by mobile diagrams. The widest system forms the upper level of the diagram, any component that is a closer double is represented as a bifurcation at next level, etc. Such diagram is simply a binary tree. We choose to encode the place of each sub-system in a mobile diagram by an integer number called level. The widest system is designated level 1, the sub-system of its primary component is level 11, and that of the secondary is level 12. If there are still closer sub-systems, they will be designated levels 111, 112, 121 or 122, and so on. This is evidently not a very efficient way to encode a binary tree, but it is easy to understand and to decode. The mobile diagram and the associated levels are illustrated in Fig. 1 (click here). Maximum number of levels found in MSC is 4, e.g. from 1 to 1111.
Level designations can be generalized to include non-hierarchical systems, called also trapezia, which contain more than two components with comparable separations. The separation ratio 3:1 is generally used to distinguish trapezia from apparently hierarchical systems. For example, a trapezium system with 3 visual components A, B, C will be represented as two binaries AB and AC, but these "binaries'' will have the same level (level 1). Should it happen that components B or C are in turn spectroscopic binaries, they would be designated levels 12 or 13.
The attribution of a sub-system to the particular hierarchical level may change with the discovery of new components. The multiple systems with complete component census (complete systems where the existence of undiscovered companions is excluded by observations and stability considerations) are exceptionally rare, if not inexistent. Hence, the terms "triple'', "quadruple'', etc. reflect the current observing status rather than the true multiplicity. The definition of a sample of complete multiple systems is of great interest, and MSC may be used as a source of observing programs needed to construct such sample.