Let us see first the time variations in the magnitude of 
, which
is related to the evolution of all the 
maser sources in ON2 and
not only to independent features. Comparison with variations of the total
flux (Lekht et al. 1996) shows that the drift jump took place
long before (about two years) the main increase of maser activity. The jump
coincided in time with an increase in activity at positive velocities.
Comparison of with the centroid time variations in ON2 (Lekht
et al. 1996), as in the case of the total flux, did not give a clear
result. Good correlation was found between the variations of
and the drift of the group of features at positive velocities (Márquez
& Lekht 1997). This correlation occurred during 1981-1988. The
second drift cycle (1990-1996) showed no correlation between parameters. At
the present time we can not try to explain the observed phenomena and their
possible causes. It will be appropriate to observe the maser in
ON2 in the decreasing part of its activity curve which began in the middle
of 1996.
The variations of 
may be interpreted as the superposition of
the slowly varying and flare components. The first of these correlates with
the long-term component of variations of the total flux and the second with
faster variations with period of about 3 years (Lekht et al.
1996). The position of the local maximum, expected at the end of
1992 -- beginning of 1993 was not well determined due to the fast increase
of the main maximum.
It was found that during any increase of the maser activity in ON2 the
parameter always increased, i.e. the acceleration of the maser
condensations occurred.
Analysis of the masers in ON2 showed that anticorrelation between
the fluxes of components with close radial velocities was not a rare
phenomenon. Below we analyze three more important cases. It should be noted
that the character of the flux variations are somewhat different in these
cases compared with that described earlier.
With competition between radiative modes for pumping in a partly saturated maser, a rapid decrease in the emission of one feature accompanies a rapid increase in the emission of arother. An example of such a situation with a repetition time of about 1.5 years may be seen in Fig. 2 (click here). This anticorrelation was consistent throughout the entire period of the existence of these features.
The flux anticorrelation between the two longest-lived features also lasted during all the time they were in the active phase (Fig. 4 (click here)). The maximum emission of one of these features always coincided with the minimum of the other and vice versa. The time delay between the two consecutive maxima lay between 2 to 3 years. The maser condensations responsible for this emission may belong to different groups, located in the front of the cometary arc (Hofner & Churchwell 1996).
The third type of anticorrelated emission appeared for the more intense
features at negative velocities (Fig. 3 (click here)). Flaring in features 34
and 33 took place consecutively. Feature 33 appeared in the
spectrum just as feature 34 disappeared. Feature 34 showed an exponential
decline and the flux of 33 after this had two different maxima. The flux
decrease of this feature was fast and almost linear over about one year.
This time dependent behaviour of the maser emission did not seem to be related to competition between modes in the two features, since it lasted throughout different maser activity phases in ON2. However, this may be a result of the reported anticorrelation between two groups of spectral features (Lekht et al. 1996).
Velocity variations of many features followed either a wave-like or an arc-like curve. Only for three features did the flux maxima and minima coincide with the velocity maxima and minima. This correlation may be interpreted in terms of the velocity increase of the condensations under the action of an external agent, such as shock waves or strong stellar winds. Following this action or its decline the deceleration of the condensations in the medium seems quite possible, with a simultaneous decrease in the emission level.
The fitting of smoothed curves may lose some small effects in the radial
velocity variations. For this reason we analyzed the differences between the
velocities of given features 
and
. This also eliminated any errors there may have been in the
determination of absolute velocities. We found that during flux increase of
any of the features, the difference between their radial velocities also
increases.
The relatively small member of features with correlation between flux and
velocity variations suggests the existence of another cause. In some cases,
it appears that superposition of the emission from neighbouring
condensations took place. Their consecutive flaring up and dying down leads
to a systematic drift in the maximum emission in the spectrum.
If the lifetime of the features does not significantly exceed the time
between the observations, either a dispersion of the dots relative the
solid curve or jumps in the velocity may appear. Such effects were observed
approximately in 15 cases.
For seven features velocity variations with a period of 
and amplitude about 
are superposed in the smoothed
curve. No correlation between these velocity variations and flux variations
was found.