Context :

Aims : Results of a long-term spectral monitoring of the active galactic nucleus of NGC 4151 are presented ( 11 years , from 1996 to 2006 ) .

Methods : High quality spectra ( S/N > 50 in the continuum near H \alpha and H \beta ) were obtained in the spectral range \sim 4000 to 7500 Å , with a resolution between 5 and 15 Å , using the 6-m and the 1-m SAO ’ s telescopes ( Russia ) , the GHAO ’ s 2.1-m telescope ( Cananea , México ) , and the OAN-SPM ’ s 2.1-m telescope ( San-Pedro , México ) .
The observed fluxes of the H \alpha , H \beta , H \gamma and HeII \lambda 4686 emission lines and of the continuum at the observed wavelength 5117Å , were corrected for the position angle , the seeing and the aperture effects .

Results : We found that the continuum and line fluxes varied strongly ( up to a factor 6 ) during the monitoring period .
The emission was maximum in 1996-1998 , and there were two minima , in 2001 and in 2005 .
As a consequence , the spectral type of the nucleus changed from a Sy1.5 in the maximum activity state to a Sy1.8 in the minimum state .
The H \alpha , H \gamma and He \lambda 4686 fluxes were well correlated with the H \beta flux .
The line profiles were strongly variable , showing changes of the blue and red asymmetry .
The flux ratios of the blue/red wings and of the blue ( or red ) wing/core of H \alpha and H \beta varied differently .
We considered three characteristic periods during which the H \beta and H \alpha profiles were similar : 1996-1999 , 2000-2001 and 2002-2006 .
The line to continuum flux ratios were different ; in particular during the first period ( 1996-2001 ) , the lines were not correlated with the continuum and saturated at high fluxes .
In the third period ( 2002-2006 ) , the H \alpha and H \beta fluxes were well correlated to the continuum flux , meaning that the ionizing continuum was a good extrapolation of the optical continuum .
We thus consider that the values of the time lags – line lagging continuum ( 0.81 ^ { +1.55 } _ { -0.81 } days for H \alpha and 0.81 ^ { +2.19 } _ { -0.81 } days for H \beta ) for the third period give a more realistic estimation of the dimension of the BLR than during the other periods . Moreover , the time lags obtained by binning intervals of three years within the whole monitoring period indicate the permanent presence of a small component of the BLR ( 0.3-0.7 light days )

Conclusions : We discuss the different responses of H \beta and H \alpha to the continuum during the monitoring period .