Context : It is well known that the tilt angles of active regions increase with their latitude ( Joy ’ s law ) .
It has never been checked before , however , whether the average tilt angles change from one cycle to another .
Flux transport models show the importance of tilt angles for the reversal and build up of magnetic flux at the poles which is , in turn , correlated with the strength of the next cycle .

Aims : Here we analyse time series of tilt angle measurements and look for a possible relationship of the tilt angles with other solar cycle parameters , in order to glean information on the solar dynamo and to estimate their potential for predictions of solar activity .

Methods : We employ tilt angle data from Mount Wilson and Kodaikanal observatories covering solar cycles 15 to 21 .
We analyse the latitudinal distribution of the tilt angles ( Joy ’ s law ) , their variation from cycle to cycle and their relationship to other solar cycle parameters , such as the strength ( or total area covered by sunspots in a cycle ) , amplitude and length .

Results : The two main results of our analysis are : 1 .
We find an anti-correlation between the mean normalized tilt angle of a given cycle and the strength ( or amplitude ) of that cycle , with a correlation coefficient of r _ { c } = -0.95 ( 99.9 % confidence level ) and r _ { c } = -0.93 ( 99.76 % confidence level ) for Mount Wilson and Kodaikanal data , respectively .
2 .
The product of the cycle averaged tilt angle and the strength of the same cycle displays a significant correlation with the strength of the next cycle ( r _ { c } = 0.65 at 89 % confidence level and r _ { c } = 0.70 at 92 % confidence level for Mount Wilson and Kodaikanal data , respectively ) .
An even better correlation is obtained between the source term of the poloidal flux in Babcock-Leighton-type dynamos ( which contains the tilt angle ) and the amplitude of the next cycle .
Further results are : We confirm the linear relationship ( Joy ’ s law ) between the tilt angle and latitude with slopes of 0.26 and 0.28 for Mount Wilson and Kodaikanal data , respectively .
In addition , we obtain good positive correlations between the normalized area weighted tilt angle and the length of the following cycle , whereas the strength or the amplitude of the next cycle do not appear to be correlated to the tilt angles of the current cycle alone .

Conclusions : The results of this study indicate that in combination with the cycle strength , the active region tilt angles play an important role in building up the polar fields at cycle minimum .