A "purge" has been defined as "exclusion from political party, government, or other institution, with or without arrest, of elements hostile or objectionable to leadership. It is also sometimes applied to mass arrest and imprisonment or execution of categories of population held to be 'socially dangerous' " (Rigby 1964, p. 566). This definition is broad enough to regard the exclusion of several individuals or a few prominent leaders as a purge. For our purposes, we favor a somewhat restricted definition, regarding as a purge only the mass exclusion or mass expulsion of members or leaders. Also, for our purposes, we are interested only in the purges of party members and leaders, not the elimination of "socially dangerous" persons outside the party structure.

Both party purges and party splits involve the loss of activists to the party. But, in the case of a party split, the loss of activists is considered to be voluntary and in a purge the departure is involuntary. A purge may be implemented by expulsion from party membership, imprisonment, exile from the country, or even death in the extreme case. We do not draw distinctions as to how the purge is carried out or what fate befalls those purged, for these types of variation are more likely associated with the political environment within which the party operates and are not thought to be indicators of party coherence.

At first glance, a purge might be regarded as an obvious denial of party coherence, for it means that strong disagreement among activists must exist to elicit the purge. But, in another light, the purging of disagreeables produces a more coherent party thereafter. Perhaps a more appropriate conception of the consequences of purges for a party is their effect on creating adherence to party policy. First, a party that has so little tolerance for disagreement that it resorts to expelling contrary members would be expected to have a relatively high degree of coherence initially. Second, the use or threat of the purge as a weapon invites members to think about the consequences of disagreement. While the absence of purges in a party is not a necessary indicator of low coherence, the presence of purges seems a sufficient condition for high coherence afterward.

Operational Definition. Instances of involuntary departure from the party during our time period qualify to be coded as "purges" only if they involve mass exclusion or mass expulsion of members or leaders. In the case of a purge of party members, we interpret "mass" to mean at least 10 percent of the total membership; in the case of party leaders, we interpret "mass" to mean at least 25 percent of the particular leadership group in which the purge occurs. The scoring matrix applied to coding BV1006 further distinguishes between major and minor purges and the frequency of purges according to the scheme in Table 11.7. A party that experienced only one minor purge in 1953 would thus be scored 1 for the first half of our period; one that experienced a major purge in 1958 and another in 1959 would be scored 3 for the second half.

TABLE 11.7 Scheme for Coding Purges
Frequency of Purge Events	Magnitude of Purge
	Minor: 10% of Membership or 25% of Leadership Group	Major: More Than 25% of Membership or More Than 33% of Leadership
No purge in either time period	0	0
One purge	1	2
Two purges	2	3
Three or more purges	3	4

Coding Results. From several standpoints, BV1006 proved to be the least satisfactory variable in the entire study. First, there was the problem of determining the proper classification for "party purges" within the conceptual framework. It was considered initially as a technique of party discipline and therefore a candidate indicator of "centralization of power." Considerable ambivalence accompanied its location under "coherence." Second, the variable was originally conceptualized without incorporating the requirement of mass expulsion or exclusion from the party, but this led to identifying as purges virtually any attempt to promote party discipline by expelling individual members for deviation from party policy. The operational definition of that conceptualization led to high scores and left coders with the nagging thought that many parties which probably also expelled members escaped detection for purges because the literature did not trouble to report expulsions in detail. Finally, the reconceptualization and new oper-

go to page 125