1: \begin{abstract}

2: Among their many uses, growth processes (probabilistic amplification),

3: were used for constructing reliable networks from unreliable components,

4: and deriving complexity bounds of various classes of functions.  Hence,

5: determining the initial conditions for such processes is an important

6: and challenging problem.  In this paper we characterize growth processes

7: by their initial conditions and derive conditions under which results

8: such as Valiant's\cite{Va84} hold.  First, we completely characterize

9: growth processes that use linear connectives.  Second, by extending

10: Savick\'y's~\cite{Sa90} analysis, via ``Restriction Lemmas'', we

11: characterize growth processes that use monotone connectives, and show

12: that our technique is applicable to growth processes that use other

13: connectives as well.  Additionally, we obtain explicit bounds on the

14: convergence rates of several growth processes, including the growth

15: process studied by Savick\'y (1990).

16: 

17: \noindent {\bf Keywords:} Computational and structural complexity, growth 

18: processes, probabilistic amplification

19: \end{abstract}

20: