Be stars have long been known to be photometric variables. These variations can have periodic and/or irregular components of different time scales: (a) periodic or multiperiodic components with small amplitudes of roughly several hundredths of magnitude in the V magnitude, characterized by short time scales to 3 days (Balona 1990, 1995) which are commonly interpreted as due to non-radial pulsations and/or to stellar surface spots or corotating features (Baade & Balona 1994); (b) periodic variations with intermediate time scales, days to 100 days, with amplitudes of several hundredths or tenths of magnitude (Harmanec & Kříž 1976) which are observed only in binary Be stars; (c) quasi-cyclic or irregular variations with rather long time scales, months, years and sometimes decades, generally implying magnitude changes that range between 0.01 and 0.4 mag (Feinstein 1970, 1975; Feinstein & Marraco 1979; Alvarez & Schuster 1981; Kozok 1985; Percy et al. 1988). The latter can, however, be as high as to 1.2 mag (Huffer 1939; Mook et al. 1974; Howarth 1979; Alvarez & Schuster 1981; Bernacca & Bianchi 1981; Divan et al. 1982, 1983; Mennickent & Vogt 1991; Apparao 1991; Roche et al. 1993; Mennickent et al. 1994; Ballereau et al. 1995). Though the irregular fluctuations are, as a rule, characterized by long time scales, these scales have been found in the case of very active Be stars to be as short as a few days (Hubert-Delplace et al. 1982). The quasi-cyclic or irregular photometric variations of Be stars are accompanied by more general spectrophotometric variations which concern a wide wavelength coverage. These variations have also been known for a long time (Chalonge & Safir 1936; Greaves & Martin 1938; Barbier & Chalonge 1941). Since Gerasimovic (1929); Chalonge & Safir (1936); Barbier & Chalonge (1939), it is also known that when line emission or "shell'' features are pronounced, Be stars can present two Balmer discontinuities (hereafter BD). The first BD looks like that of a normal B star and it has the same spectral location. The second BD appears at shorter wavelengths, which means that it is formed in a medium with a lower pressure than in the photosphere of the central star. It can be in emission (during a strong emission line phase) or in absorption (during a strong "shell'' or absorption line phase). The value of the second BD in emission not only depends on the spectral type (Peton 1981) but is correlated with Balmer emission line strength (Schild 1978; Feinstein & Marraco 1979; Peton 1981; Kaiser 1989; Dachs et al. 1989) and with Balmer decrement as well (Divan et al. 1982). The emission component of the BD is generally not higher than -0.15 dex, but may exceptionally be as high as -0.25 dex (Barbier 1947). When the second BD is in absorption, it can sometimes represent a flux deficiency stronger than 1 mag (Zorec 1986, 1994). The second BD disappears during a faint emission/shell line phase, so that the stellar continuum energy distribution near the BD looks like that of a normal B star. During the spectrophotometric variations of Be stars, the first BD maintains the value and mean spectral position measured during an emissionless phase to closer than 0.02 dex and Å respectively (Divan & Zorec 1982a,b). The constancy of the first BD to within 0.03 dex was also confirmed by Dachs et al. (1989) and it is currently used for spectral type classification (Divan 1979; Divan & Zorec 1982c; Divan et al. 1982, 1983). Therefore, it is also useful to determine the effective temperature of the central object (Divan & Zorec 1982c; Zorec et al. 1983; Zorec 1985, 1986; Kaiser 1989; Dachs et al. 1989; Zorec & Briot 1991) and the surface gravity (Divan & Zorec 1982a; Zorec 1985, 1986).
It has been shown that the irregular photometric variations imply a global continuum flux excess or deficiency as compared to stellar underlying photospheric radiation. This can be estimated as , where (V magnitude of the star+envelope system; V* magnitude of the star alone) and it is for continuum emission (or flux excess) and for continuum flux deficiency. The magnitude excess is positively correlated with Balmer line emission strength (Zorec & Briot 1985, 1991; Dachs et al. 1988, 1989; Ballereau et al. 1995) and with reddening of the Paschen continuum (Divan et al. 1978; Divan 1979; Chkhikvadze 1980; Dachs 1982). In turn, this reddening is correlated with Balmer line emission strength (Dachs et al. 1988). The visible flux deficiency is generally small ( mag) and it is not accompanied by marked changes of visible Paschen energy distribution (Divan & Zorec 1982c; Zorec et al. 1989; Zorec 1994). Quasi-cyclic and/or irregular changes from positive to negative flux excesses are symptomatic of the known Be "phase'' variations Be B-normal Be-shell. The present paper is devoted to the irregular spectrophotometric variations of Be stars.
The existing observational studies on long-term photometric and/or spectrophotometric variations of Be stars can be classified into two broad groups: (1) studies where authors used observations of their own; (2) studies based on compilations of published observations each made in a given photometric or spectrophotometric system.
(1) - The most relevant studies in the first group are those of Pavlowski et al. (1996) for northern Be stars and Manfroid et al. 1991, 1995; Sterken et al. 1992, 1993 for southern ones. Observations presented by Pavlowski et al. (1996), done in the UBV system, concern the monitoring of 76 Be stars over 18 years since 1972. In this program there are 28 well-observed stars with 9+8-3 observing seasons and each with 392+805-288 individual observations, and 48 less observed objects with 3+2-2 observing seasons each and with 60+159-57 individual UBV measurements. Manfroid's et al. 1991, 1995 and Sterken's et al. (1992, 1993) 12-year observing program also concerns, among other variable stars, 15 Be stars which were monitored in the uvby Strömgren system since 1982 for about 6 months per year and with a time frequency between 2 and 14 days. From Pavlovski's et al. (1996) study it appears that the number of stars with variations on time scales from months to years is about the same as the number with time scales of days or fractions of days. These variations are, however, 3 times more frequent among stars earlier than B5 than in the later ones. Six stars present orbital components of photometric variations and only 2 show sudden light changes. From the published light curves, which in many cases reveal complicated variations difficult to summarize, we can distinguish those with shell-type characteristics, where two dominant behaviours seem to appear: (i) those of stars where the (U-B) colour index changes much less than the (B-V) index; (ii) those, the most numerous, where the (U-B) colour index changes much more than the (B-V) index. Both types of behaviours are seen in early as well as in late type Be stars. Mennickent et al. (1994) report a detailed study on 13 Be stars based on uvby Manfroid & Sterken's et al. 1991-1995 observations. These authors note: (a) quasi-periodic oscillations on time scales between 10 and 20 days with amplitudes up to 0.2 mag preceding fading events; (b) in 7 stars, which mostly have shell characteristics, they observe that: (b1) small or rather no variations of the b-y index accompany the variation of magnitude u; (b2) there is a correlation between index c1 and magnitude u; (c) in all 13 studied Be stars and for all uvby magnitudes, there is, within the uncertainties of coefficients defining the regression lines, a mean relation between the long-term variation amplitudes and the standard deviation of short-term variability. Concerning the (c1,u) relation, it is worth noting that in shell Be stars not only the variation of u is noticeably higher than in v but also , so that the correlation between c1 and u seems natural. However, the well-defined trend between the slope of the (c1,u) relation and , which identifies possible differences in the relative behaviours of Balmer to Paschen continua, could still deserve further attention.
The first group also comprises studies of long-term spectrophotometric variations of Be stars made using the BCD system (for definition of the BCD (Barbier-Chalonge-Divan) system, see Chalonge & Divan 1952). The great advantage of this system is that it reflects the variation of continuum energy distribution only, because it is not affected by spurious "reddening/blueing'' effects introduced by variable spectral lines entering the wide wavelength intervals of some photometric filter pass-bands. The BCD data were obtained at irregular time intervals over more than 40 years. They concern a small number of Be stars, of which most are of early spectral types. These data led Divan (1979); Divan et al. (1982); Zorec (1986) to conclude that in the studied objects the continuum emission phases were characterized by well-defined relations between the visible gradients of energy distributions, total Balmer discontinuity and visual magnitude. These relations seem to be different for each star and unchanged even after the star had a phase of apparent loss of emission characteristics. These authors also found that, in general, reddening of the visible gradient is positively correlated with stellar brightening. Inversely, "shell'' phases are characterized by very small changes, if any, in the visible energy distribution.
(2) - In the second group are papers concerning either a small number of objects at typical "Be'' phases, or a large number of objects at any variation phase. In the first subgroup, works like Dachs (1982); Dachs & Hanuschik (1984) and Dachs et al. (1988) show that long-term photometric variations can be characterized by slopes ranging between about -0.1 and -0.3 for individual stars and that reddening of the (B-V) index is related to stellar brightening. In the second subgroup, it is worth mentioning Nordh & Olofsson (1977); Hirata (1982) and Hirata & Hubert-Delplace (1981). From UBV data of 50 to 70 Be stars spread over more than 10 years, Hirata et al. concluded: (i) that early type Be stars get redder when they brighten, and on the contrary late type Be stars are bluer when they brighten; (ii) the slope is steeper when is higher, and that this slope has the same sign as the colour excess ratio .
Finally, there are some attempts of a theoretical nature to explain the long-term photometric variations of Be stars. Using a cylindrical model for the CE, Hirata (1984) concluded that changes of V and (B-V) are due to variations of the extent of the circumstellar envelope (hereafter CE) and that is a function of spectral type and aspect angle. The irregular variations of V magnitude were also discussed by Zorec & Briot (1991). They concluded that when , the irregular variations may correspond to two different opacity regimes of the circumstellar envelope: (a) a low opacity regime, where variations of V are mostly due to small changes of opacity, and (b) a high opacity regime, where variations of V are mainly due to changes of the extent of the CE. Inversely, for there is only a high opacity regime. In a recent work, Hirata (1995) studied the long-term variations of some Be stars at shell phases and concluded that their variations are produced by extended photospheres.
As continuum and line variations correlate only when strong line emissions or line absorptions are present during "spectroscopic'' Be and Be-shell phases respectively, to avoid misunderstandings relative to spectroscopic or spectrophotometric variations, whenever necessary we explicitly refer in this paper to "spectrophotometric Be phase'' (hereafter SPh-Be) when the second BD is in emission and to "spectrophotometric shell phase'' (hereafter SPh-shell) when the second BD is in absorption.
Results presented in the preceding section are consistent and complementary to each other. The relations obtained between photometric colour indices and/or spectrophotometric parameters show that they should somehow depend on the physical structure of the CE and on its changes. In particular, early BCD results suggested that these relations are permanent for each star and differ from one object to another, as would be the case if there were permanent characteristics related to the structure of the CE of Be stars which subsist even after phase changes or apparent losses of emission. However, in addition to opacity and geometrical factors cited in the preceding section, still other phenomena could be responsible for photometric/spectrophotometric changes of Be stars. In fact, as the contributing regions in the CE to the variations are expected to be located rather near the underlying star (Zorec & Briot 1991; Zorec & Garcia 1991; Cruzado et al. 1994) ionization/excitation balance changes in the deepest layers of the CE (Magnan 1979), as well as radiation transfer effects related to non-thermal phenomena (Gebbie & Thomas 1970, 1971; Thomas 1983) should also be relevant. Due to all these effects and because some of the above observational results were obtained from a small number, mostly early type Be stars, it would still be necessary: (a) to analyse a larger sample of Be stars of all possible subspectral types, each having been observed as frequently as possible and over a long time base; (b) to study the actual character (constancy, uniqueness, etc.) of relations among spectrophotometric parameters for each star, especially, if feasible, for those having undergone phase changes. This study aims at giving as far as possible comprehensive observational bases on long-term spectrophotometric variations of Be stars, derived from compiled photometric and spectrophotometric data obtained in various systems, to ensure a large time base coverage and for a quite a large number of stars.