1. Introduction

Central to most treatments of the dynamics of voting is the assumption that the voter selects whatever party, candidate, or issue alternative he judges to be the least "distant" or the most similar to his own position or ideal preference. Such an assumption obviously underlies both theoretical work and journalistic observation, and is quite explicit in such relatively formal models of the voting process as those proposed by Anthony Downs¹ or Davis and Hinich.² Indeed, from some points of view it is difficult to imagine what kind of contrary notion would have any plausibility whatever.

While it is quite natural to depend on such an assumption, it is rather more difficult to say what one means by "distance" or "similarity" in a more operational way, in order to permit empirical investigations of such judgments. In an earlier paper, one of us has provided an account of some of the properties of distances perceived by voters in France and Finland as separating the multiple political parties of their respective systems within some multidimensional space.³ These perceptions were deduced from sets of preference orders for the various parties of each system provided by samples of the citizenry.

This investigation indicated that citizens within each of these countries displayed quite heterogeneous assessments of party location, which were marred further by frequent ignorance of the existence of some parties and broad "indifference zones" covering sets of other parties. Thus any expectation of a clear and consensual location of parties on some single dimension of ideology or policy controversy is not at all realistic. However, by averaging the data to highlight central tendencies,

---

* The authors wish to express their thankfulness to Mr. Sten Sparre Nilsson who has read the manuscript and made valuable suggestions. However, the authors take the full responsibility for the content.

* The authors wish to express their thankfulness to Mr. Sten Sparre Nilsson who has read the manuscript and made valuable suggestions. However, the authors take the full responsibility for the content.

expanding the analytic model to multiple dimensions, and taking into account the fact that different dimensions are important for different voters, the configuration of distances thereby revealed came to illuminate in a very plausible way the perceptual organization of these spaces in the mind of the electorate.

Among other things, the data presented gave indication that these perceived spaces had some Euclidean properties, at least in a crude sense, instead of following certain reasonable but more primitive alternative models. There seemed to be meaningful differences between voters, moreover, with respect to the sheer size of the perceived party system: politically uninformed and uninvolved citizens saw distances between parties as being smaller than they were considered to be by their more involved compatriots. The hypothesis that specific dimensions defining the space of party competition would be perceptually exaggerated or foreshortened according to the relative importance of the dimension for the voter, was also confirmed as far as the data could be carried. Finally, at a more descriptive level, it was possible to reconstruct an average "picture" of perceived party locations in Finland and France. This work suggested that there were rather remarkable similarities in the clustering of analogous parties in the two systems as perceived by their voters. All told, the results seemed to show the great utility of such reconstructions in the analysis of major dimensions of political cleavage in multiparty systems, as well the evolution of such cleavage over time.

The earlier paper was hampered throughout, however, by the fact that the data have been collected in both Finland and France without such multidimensional analyses in view, and hence were not ideally adapted to them in a number of respects. In the interim, more appropiate data have been collected in several multiparty systems, and analytic methods for the treatment of multidimensional similarity data have enjoyed a considerable flowering as well.

In particular, recent sample surveys of the Norwegian electorate during the 1965 and 1969 national (Storting) elections have included information collected more expressly for these purposes.⁴ Furthermore, the Norwegian site is especially congenial for the analysis of cleavage dimensions, since perhaps more has been done in Norway than in any other multiparty system toward the historical reconstruction of cleavage lines evident in popular voting.5

Therefore this paper is a frank sequel to earlier work by both authors. Its purpose is in part to use the new data to add the perceived party configurations for Norway to the earlier ones from Finland and France. More important, however, we wish to capitalize on the more adequate data to extend the initial work by a number of steps. Most notably, we are anxious to explore differences in the perceived shape of party locations across subpopulations of the society, whether defined in social or psychological terms. We also wish to consult the 1969 data for evidence of stability and change in these configurations over time. First, however, it is important that the reader be introduced to the numerous parties which currently compete for the votes of the Norwegian citizen, as well as the shifting lines of cleavage which spawned them in the past.

2. The Cleavage Structure of Norwegian Politics

The chief axes of policy dispute in a nation at any point in time are probably restricted in number, if for no other reason than that of the limits of attention on the part of either the governors or the governed. However, for most countries most of the time, such distinct axes number more than one, and they tend moreover to wax and wane in their urgency and salience in a complex interplay as time passes. Changes in social or economic structure, as well as shifts in the foreign situation, create new cleavages or resuscitate old ones, while drawing attention away from still others. And it is against such a pattern of change that political parties may fractionate, arise anew, suffer serious reverses, or die.

Shifts in lines of salient cleavage do not necessarily produce changes in the constellation of political parties in a system, although they will certainly change the terms of their competition. Whether old parties disappear and new ones emerge depends, inter alia, on the way the masters of the old parties accomodate to new definitions of the situation and the degree to which new disputes fall along old cleavage lines or cross-cut them. Moreover, the probability that new parties will be established at a time of cleavage change is strongly conditioned, as many have pointed out, by political institutions in general and the electoral system in particular.6 6 For Norway, the evolution of party history over the past century and a half is sufficiently complex and varied to assure us both that the electoral system, which itself has undergone dramatic change, has generally been conducive to party formation, and that a fair degree of change in lines of cleavage has occurred.

Representative government was introduced into Norway by a new constitution in 1814, and almost immediately tendencies toward party grouping began to appear in the Parliament (Storting). Representatives of farm background joined against the urban civil servants of the upper class who dominated the government and were to retain their hold on it over much of the century. At first, this opposition to government was fluid and inarticulate. It gained strength, however, after 1830 as a radical democratic movement began to emerge in the cities, and gradually allied itself with the rural opposition. By 1870 this coalition was increasing its demands and coming to behave more and more like a political party. It was called Venstre (the Left), whereas supporters of the regime were called Høyre (the Right).

The first straight partisan elections were held in 1882, although formal nationwide parties were not established until the introduction of parliamentary rule in 1884. In these earliest years of the modern period, the most salient lines of cleavage tended to be of two related types: the territorial (T) and the socio-cultural (SC).⁷ It was a conflict on the one hand between the capital and the provinces, the center vs. the periphery. It was also a socio-cultural conflict between the academically educated, "Europeanized", patrician officialdom of the cities and the increasingly status-conscious, articulate, and politically-oriented farmers of rural districts and their sons in the expanding cities.

The Left did in fact capture power in 1884, but was able to remain a united

party for only about four years. At that time, a new cleavage became salient, and the Left split into two camps: a "Pure" wing of radical nationalist, and a "Moderate" wing of spokesmen for traditional religious and moral values (R). The Moderates had their strongholds in the coastal areas of the south and west, while the "Pure Left" was dominant in the interior fjords and valleys. In this new configuration of three parties, a temporary alliance arose between the conservatives at the nation's center, and religious fundamentalists at the periphery.

Soon, however, the threat of war over the union with Sweden drew attention from the domestic opposition of Left and Right, at much the same time as universal male suffrage brought the growing class of workers into the electorate. Thus the stage was set for a series of changes in party alignments. The Labor Party had in fact been established as early as 1887, although its partisan political energies were channeled for some years through the existing Left. After the turn of the century, however, with its new resource of workers' votes, it rapidly rose as a prominent party in its own right, becoming the dominant party in the northern periphery ⁸ Another left-wing party, the Worker Democrats, had also been established in the 1890's appealing to forest workers and tenant farmers in the eastern region of the country.

Faced with the competition from these two new parties, the original Left decided to radicalize its policies. It demanded a number of social reforms, more state control of the economy, and legal protection of Norwegian natural resources against exploitation by foreign as well as domestic capitalists. Thus while the earlier lines of cleavage had displayed some familiar features of opposition between the "haves" and the "have-nots", this early period of industrialization began to produce a new and potent cleavage centered more squarely on the labor market (LM).

Termination of the union with Sweden (1905) permitted many of these domestic issues to return to the fore. A right wing of the Left party, put off by the radicalization, split away to form a new party (the National Liberals), or to join the Conservatives. At the same time, the fundamentalist Moderates either came to feel that other issues than religion were more important or found Liberal policies on religion more to their liking, and coalesced again with the parent party. In the same years before World War I, the character of the Right shifted gradually under these political pressures from being a party of officialdom to one with more emphasis on defence of business, industry, and shipping from interference by the state.

The rapid progress of industrialization after 1905 exacerbated class antagonisms, which became the more acute because of inflation and speculation during World War I. The radicalization of the Labor Party, originally quite moderate, continued. A syndicalist wing modelled partly after the American IWW was formed in 1911, and in 1919, strongly stimulated by the Russian Revolution, the party joined the Communist International. At this point a more moderate minority broke away to form the Social Democratic Party. Soon, however, finding difficulties in coping with Moscow, the Labor Party left the International (1923), with a minority splitting again to maintain membership and become the Communist Party. In 1926 the Social Democrats rejoined the Labor Party, and the unified group won 37 % of

the voters in the national election of 1927. From that point on it was the largest
party in the system, while the Communist Party, still antagonistic to it, rapidly
dwindled to relative insignificance.

Unrest characterized all the parties after World War I, at the same time as the introduction of proportional representation (1920) gave minor parties a greater chance of winning seats in elections. Emotions were such on the socialist left that fractionation would probably have occurred whatever the electoral system. However, the Liberals of the older Left, now rather centrist in the system, found themselves split asunder by a new cleavage dimension, the conflict between rural and urban interests in the commodity market (CM in our summary, Table I, below). The main farm organization (Landmandsforbundet] had run a few candidates in the elections of 1915 and 1918, and from these roots, along with the facilitation of proportional representation, the Agrarian Party was formed in 1921.⁹ This new party sapped away much of the remaining strength of the old Left in the rural areas of the East Inland and Trøndelag, while making considerable inroads as well in the agricultural areas of the Southwest and North.

Similarly, we may conjecture that proportional representation facilitated the birth of the next party in the system. A highly controversial prohibition law had been passed in 1919, but a government controlled by the Liberals arranged a 1926 referendum repealing the law. The numerous fundamentalists within the ranks of the Liberals became suspicious the party was no longer able or willing to defend their interests, and began to look elsewhere. In 1933 the Christian People's Party was formed for competition in some western provinces, and in 1945 it established itself as a national party. Thus the religious-moral cleavage (R) had again intruded to fractionate the party system. A comparison of the geographical basis of the Moderate Left in the 1890's and the Christian People's Party after the Second World War shows a remarkable similarity. Both parties enjoyed relatively strong support in coastal areas, particularly in the western region.₁₀

The 1920's and early 1930's were an unstable period in Norwegian politics more generally. No party at this time enjoyed a majority in the Storting. The Left and Right and later the Agrarians alternated in power. Only minority cabinets seemed possible, and crises of government were frequent. From the long historical view it may seem strange that parties of the Right failed to form stable coalitions during exactly the period when they were faced with the most aggressive and revolutionary surges of the labor movement. However, this fact speaks for the lingering strength of traditional antagonisms between the old Left and the Right, exacerbated again by a return to salience of moral-religious issues as well as the question of language reform, which touched the scars of the old socio-cultural cleavage.

In 1935 the Labor Party, by then a more moderate reformist group, formed a minority government lasting until World War 11, and from 1945-61 enjoyed a majority position in the Storting as well. This period was a relatively quiescent one for the composition of the party system. In 1961, however, the last of the current seven political parties was formed as the Socialist People's Party. Again, a new line

of cleavage had intruded. In 1948 Norway had joined the Western Military Alliance on the broad agreement of the Labor Government and all of the bourgeois parties. Only the Communist Party and a left wing of Labor opposed the move. In the wake of the Cold War, this neutralist feeling in the ranks of the Labor Party led to the creation of the Socialist People's Party, which demanded that Norway leave NATO and which took a stand somewhat to the left of the Labor Party itself in economic and social matters. In the election of 1961, the new party won two seats in the Storting and took away Labor's majority for the first time since the War (Labor was reduced to 74 seats, against 74 held by the four bourgeois parties jointly). For most of the next four years Labor ran a minority government, but in 1965 the bourgeois parties won 80 of the 150 seats in the Storting and organized a coalition government. At the election of 1969 the majority of the bourgeois parties was reduced to a minimum: 76 bourgeois versus 74 laborites, whereas the Socialist People's Party lost its two mandates. The bourgeois coalition government remained in office after the election, but had to resign in March 1971 due to internal struggles. The Labor Party formed a minority government.

From this brief survey it should be clear that the formation and disappearance of parties in Norway has evolved around the five major cleavages summarized in Table 1.11 These cleavages have developed gradually, and there are several indications that the salience of the various cleavages has shifted over time. The moralreligious conflict, for example, twice erupted to produce new fissures in the party system. For more than three decades in between, however, the capacity of the Left to integrate the fundamentalist interest in its program, along with attention drawn in other directions, meant a period of reduced salience of these concerns within the party structure.

Of equal importance for our purposes in this paper, is the recently developed capacity for cooperation among the bourgeois parties. The cleavages splitting the bourgeois ranks appear to have been too strong to have allowed such a coalition in the 1920's despite the confrontation with a revolutionary Labor Party. In the period after 1945, bourgeois cooperation would have been the only way of creating a viable alternative to the Labor majority, but traditional antagonisms still prevented the four parties from establishing joint policies save in a few minor areas.¹² It was not until the 1960's that the four bourgeois parties were able to join forces in presenting a government alternative. It seems plausible to imagine that the traditional conflicts which separated the four parties have become significantly less salient as time has worn on. While we lack the data to measure the relative strength of various cleavages in any precise way for the historical past, we shall be particularly interested in the way our estimates of party distances perceived by electors behave in the period from 1965 to 1969 for the bourgeois parties.

Finally, it is important to keep in mind that the distances which voters perceive to separate a set of competing parties at any point in time is not exclusively a function of the relative salience of various cleavages at that moment, but also of the degree to which the different substantive disputes intercorrelate or cross-cut one another. Such intercorrelations tend to exist jointly at both the level of voter

attitudes and the level of party programs. In the nineteenth century, the territorial and socio-cultural conflicts were fought out in different terms, and in some degree involved different interest groups in the system. Nevertheless, the two cleavage lines scarcely ran at right angles to one another. Voters in the periphery struggling against the dominance of the center tended at the same time to have less of the European academic background that marked the culture of the cosmopolitan center. Similarly, the two major proto-parties of the Left and the Right reflected these correlations, lining up as "haves" and "have-nots" on both dimensions, such that the gulfs between "left" and "right" were being doubly reinforced. The main market antagonisms of the later period, while again setting rather distinctive groups (such as farmers in one case and urban labor in another) against one another, continued to produce cleavages sufficiently correlated with the old Left-Right battle lines to lend some reinforcement to this underlying dimension. Indeed, of the five cleavages we have distinguished, it is probably the moral-religious conflict which has most directly and frequently cross-cut the others, a matter admirably reflected in the fact that first the left wing and then the right wing have seemed the most nearly congenial "homes" for those voters whose interests in policy competition are predominantly fundamentalist. Moreover, it would be possible to distinguish at least two policy arenas within the moral-religious cleavage (prohibition and the question of religious instruction in the schools), but the partisans of both overlap so completely that it does not seem worth making such a differentiation.

3. Current Reflections of the Historical Cleavage Structure

Although the intensity of particular cleavages may wax and wane over time, constancies in public attitudes on such issues, along with the firmness of party allegiances once developed, means that it is not hard to find in current data some cumulative, almost geologic, record of the past. Before proceeding further it is worth displaying parts of this record, which show not only the durability of such cleavage lines, but also the fact that while a number of cleavages are correlated roughly with the classic "left" and "right", they are somewhat distinctive, generating in some instance truly triangular conflict patterns.

The territorial conflict, one of the oldest in the system, still shows itself clearly even in aggregate voting returns.¹³ The Conservative Party maintains its traditional stronghold around the Oslofjord, and is much more likely to draw its votes in other regions from the cities than in rural areas. The three middle parties (Liberals, Agrarians, and Christians) are most popular in the Southwest, and their support outside that region tends to be rural, although most markedly so, of course, for the Agrarians. The three socialist parties tend to be slightly stronger in urban than in rural communities. In the Southwest, they usually gain less than 40 % of the vote, but hold a solid majority elsewhere.

It is the pair of economic cleavages surrounding the labor market (employer

vs. employee) and the commodity market (producers vs. consumers of products from the primary sector) which, while not totally uncorrelated, do produce a very clear triangular set of oppositions. These are quite apparent if we examine the way in which party voting differs within occupations (Table II). Three occupational categories are particularly homogeneous in their party preferences: (1) manual laborers, who vote mainly for the socialist parties; (2) self-employed businessmen and professionals, who are strongly inclined to vote Conservative; and (3) farmers and forest owners, who tend to support the Agrarians. Each cleavage has clear elements of opposition between the "haves" and "have-nots", one in the urban status hierarchy and the other in rural areas. But some conflicts of interest between functional sectors intrude to prevent a simple bipolar opposition. The Conservative Party is most attractive to people of higher status in the cities, and the Agrarian Party to their counterparts in primary production. Labor finds itself opposed to both of these parties through its defense of the wage-earner. But Labor and the Agrarians have also jointly opposed the Conservatives in efforts to protect the primary economy against exploitation by the urban bourgeoisie and competition from abroad.

Even Table II fails to express the full lines of economic conflict, for it ignores some important regional variation in party support. Thus, for example, while Labor is generally the champion of the underdog, it has gained little support among smallholders and fishermen of the Southwest. Moreover, the table largely ignores the effects of other cleavages. The socio-cultural opposition, and even more especially the moral-religious conflict, cut quite markedly across these economic divisions.

While Table II makes clear that sympathizers of the various parties differ in their occupational composition in ways that would fit historical cleavages, it does not in itself constitute proof that those voters perceive discrepancies along economic lines in the current programs of the respective parties. However, it is not hard to add such a demonstration. One of the liveliest issues currently dividing the Left and Right involves how steeply progressive the system of taxation should be, and it is the Conservative Party which has been most vehemently opposed to the uncommonly steep progressiveness of tax rates instituted by the Labor Government over the period from 1945 to 1965.

Attitudes toward progressive taxation were measured among respondents in the 1965 survey.¹⁴ Furthermore, in the vast majority of cases where people favored one side or the other, they were also asked which party or parties were closest to their own opinion on the matter. We have divided the relatively stable party supporters in our sample according to the position they chose on the taxation item, and then examined their perceptions of which parties are closest to them in this regard. The results are presented in Table 111. There is some natural tendency, visible in the tabulation, for voters to perceive their own party as being closer to their preferred positions on the issue. However, these trends are thoroughly overwhelmed by what observers would consider to be "accurate" perceptions on the issue: the Labor Party and the Conservatives are most commonly accorded the appropriate polar positions

in the debate, even among sympathizers of other parties. Thus the current lines of
economic conflict are indeed very clearly perceived.

Where the socio-cultural conflict is concerned, the language problem has been a central issue for a long time. The old Norwegian language had been replaced by Danish as the official language during the long period of union with Denmark (1389-1814). After the new and independent constitution the urban officialdom, trained exclusively in Copenhagen, continued with written Danish, which city people in the East and in Bergen had no difficulty understanding. However, the discrepancies between this official written language and the spoken dialects were very marked, particularly in the hinterland valleys and in the West. By midcentury a gradual language reform was under way in the cities, as increasing numbers of Norwegian words and phrases were accepted in the official language, producing the distinctive urban riksmål.

No such easy accommodation of the written and spoken languages was possible in
the countryside, especially in isolated regions. The rural population therefore had
great difficulties communicating with officials and city people. A farmer's son, an

autodidact of genius, Ivar Aasen, constructed a standard rural language, the landsmål, after close study of many rural dialects. This language gave a majority of farmers a new cultural identity, a form of communication they could regard as the nation's own. It became a rally ing-point for rural opposition in the nineteenth century and has lingered on as an issue in this century. For one thing, the landsmål users have demanded that their language should be given an equal status with riksmål as an official language, e.g. in public schools and in public administration. On the formal level equality has been recognized since 1885, but the landsmål camp has always claimed that equality between the two languages does not exist in reality. Second, it has been demanded by a substantial wing of the landsmål movement and also by many riksmål users that riksmål be made more "Norwegian" in its grammar and vocabulary to narrow the gap between the written and spoken language. But stiff resistance on the riksmål side has been raised nonetheless. Over time, progressive language reforms have indeed reduced the discrepancy between the languages, but conflict over still greater convergence remains.

In view of this dispute, it is of interest to know whether the sympathizers of the various parties still differ in language use. The 1965 survey asked people which language they preferred for writing, as well as how interested they were in the language problem generally. A combination of these items produces the index in Figure 1. The graph shows how much the voters of the several parties deviate from the national average for language preference and involvement in the problem. It is clear that adherents of the bourgeois parties are much more likely to be concerned about the language problem than are sympathizers of the left, and within the set of bourgeois parties there is a clear polarization on the issue as well. A preference for landsmål characterizes sympathizers of all the middle parties, and Liberal voters are most likely to be involved in the question. The Conservatives represent the other extreme: voters there are strongly involved and overwhelmingly in favor of riksmål. Language preferences of the less involved socialist voters simply lie near the national average. These patterns mirror a good deal of history: the Right and the old "Left" (now the Liberals and the other two middle parties) were always the main antagonists in the language conflict.

Again, however, it is of interest to know whether these correlations are mere historical remnants, or whether party alignments on the controversy are still fresh in the voter's minds. In recent decades the predominant political formulation of the conflict has been the following: should the authorities enforce an amalgamation of riksmål and landsmål, or should the two languages be permitted to develop independently, retaining their individuality? People favoring a fusion of the language into samnorsk have been found in both language camps, but the most vigorous fight against such a language reform has always been waged by riksmål users.

In the 1965 survey, respondents were also asked to choose between samnorsk and a free language development, as well as the party that came closest to their own opinion on this aspect of the language question. Of Conservative supporters, whom we have already seen in Figure 1 to be almost exclusively riksmål users,

only 9 percent favored a language fusion. Across the other parties, differences in attitudes toward samnorsk are small (between 24 percent for the Christians and 34 percent among the Liberals and Agrarians). If we group the stable identifiers of the various parties according to their language preference (riksmål vs. landsmål] and ask what parties they cite as closest to themselves on the language question, we arrive at Table IV.

This table has quite a different cast from that of Table 111. For one thing, people are much less inclined to see the language question as a controversial issue. Roughly two out of three in both language camps find the issue nonpartisan. Moreover, party supporters display much less clear perceptions of the parties closest to themselves in the language conflict. Riksmål users, even outside the Conservative Party, cite that party favorably with some frequency, but this tendency remains quite weak. Similarly, on the landsmål side there is some trend toward choice of the Liberal Party as closest, but the selection is irregular across parties. All told,

positioning of the parties is much less clear in the public mind than we found to
be the case for the taxation issue, despite the fact that the composition of the
parties by language use remains fairly distinctive, particularly on the bourgeois side.

Several specific issues have contributed to the moral-religious conflicts. One of the older issues that, as we have seen, contributed to the birth of the Christian People's Party, was the enactment and repeal of prohibition. We collected data in 1965 which permitted us to distinguish between teetotalers and those who drank at least occasionally, further differentiating the former by interest in temperance and the latter according to their attitudes on numerous measures which still exist controlling the use of alcohol.ls We also collected information on religious participation which permits the construction of another index obviously relevant to the same general dimension.l6

Temperance preferences are of course strongly correlated with religious activity,
and whether we look at voting support of the parties in terms of temperance practice
(Figure 2) or religious fidelity (Figure 3), the patterns are extremely similar.

Labor voters reflect quite closely the average picture for the nation as a whole, although extremes tend to be underrepresented. The middle parties tend to show clienteles that are high on both temperance and religious activity, with the Christian People's Party quite appropriately the most extreme case. Both the Conservatives to the right and the Socialist People's Party to the left have a marked underrepresentation of these moral-religious emphases. Thus we find that there are noteworthy differences between party supporters in these terms, although it is clear that the largest discrepancies are concentrated within the bourgeois parties.

While some political issues that have been central to the moral-religious cleavage are no longer prominent, others remain quite controversial in the current period. A good example is the question of religious instruction in the schools. In view of the Lutheran State Church in Norway, there is almost complete consensus favoring religious instruction in the schools. But there is frequently spirited debate as to how closely tailored to the doctrines of the official church such instruction should be, and over the amount of attention to be given to it in the school curriculum.

This was the main form of the controversy at the time of our 1965 study. Shortly before the election of that year the Labor Government had proposed a new "normal plan for schools", whereby two hours a week on the average would be given over to religious instruction. But fundamentalists immediately demanded that school authorities be permitted the local option of extending religious instruction by one or more hours. The 1965 sample was asked whether it favored the local option amendment, or preferred to limit religious instruction to the original plan. They were also asked which party or parties they expected to stand closest to their own views in the upcoming Storting debate on the bill.¹⁷

These party perceptions on the issue by party of own identification are displayed in Table V. If we consider first the marginals at the bottom of the table, we see that the rank and file of the various parties differ widely in their own views on the controversy. Absolutely all identifiers of the Christian People's Party favor the local extension of religious instruction, and most Agrarians (89 %) want it as well. At the opposing extreme, only 32 % of Communist and Socialist People's adherents are favorable to the local option. Perhaps most interesting are the views of Labor identifiers. Despite the fact that it was a Labor Government which had made the original proposal, the party's identifiers are almost exactly split over the amendent.

The interior of Table V makes clear that the Christian People's Party is broadly perceived as closest among people wishing extended hours. Even outside of the Party's own rank and file, the mode of each distribution in the left half of the table selects the Christian People's Party, with own party being relegated to second place. However, the picture is different among those who do not wish extended religious instruction, as shown in the other half of the table. No single party is commonly selected as articulating such a position, and all partisans most frequently end up choosing their own party. A parallel asymmetry between the two halves of the table occurs with regard to the proportions of people seeing religious instruction as controversial: a lower proportion of people unenthusiastic about the extension makes party links than is true among supporters of the local option. All told

then, the picture is very clear. The Christian People's Party is widely seen as representing the moral-religious pole of the dispute. However, there is a good deal of ambiguity concerning spokesmen for the opposition, and probably a lesser salience of the issue generally among those not interested in the extension of religious instruction.

By way of summary, then, we have found in current data from 1965 a strong reflection of the succession of cleavage lines that were important in shaping the Norwegian party system. We have examined evidence from two levels: first the more or less "demographic composition" of 1965 voters for the six or seven parties; and secondly, the perceptions of party identifiers concerning which parties in the system are the prime advocates of polar positions with regard to current debates springing from these underlying cleavages.

The evidence at the first level was quite homogeneous. Although the absolute size of discrepancies in background composition of party support differed somewhat from cleavage to cleavage, in all cases we found very clear differentiation in support that fits with what is known about the historical generation of the parties. Whether voters are divided by region, occupation, linguistic preference, temperance habits, or religious activity, the expected differences in party clientele emerge unequivocally where they should. However, in view of the long-term stability of partisan preferences, as well as the parallel stability of the demographic traits or habits, results at this level might well have significance which is primarily "geological". It was in order to assure ourselves of the current vitality of these cleavage lines that we turned to the second level of evidence.

We should not miss the fact that there was a very tight fit between results from the two levels. Thus, for example, the materials on perceptions of party positions on the taxation issue suggested a strongly bipolar situation with the Conservative Party forming one extreme and the socialist parties, particularly the highly visible Labor Party, forming the other, just as high-status employers were found to be concentrated among Conservative voters arid the broad mass of employees were concentrated in the socialist camp. Or again, where current forms of the language conflict were concerned, there was a less marked bipolarity, with Liberals and Conservatives taking up the extremes, just as the Liberals showed the highest concentration of landsmål users, and riksmål preference was heavily concentrated within the clientele of the Conservative Party. Similarly, while there are fairly wide differences across the parties in the moral-religious composition of their support, the Christian People's Party stood out as a very clear - indeed, essentially unanimous - extreme. And in the perceived party position materials, a more unipolar situation emerged as well, with the Christian People's Party seen as the main agent for extended religious instruction, and none of the other six parties standing out as an articulate opposition to moral-religious demands. While there may be clear resistance to these demands scattered throughout the other parties, none of these parties seems prepared to launch the kind of crusade for the disestablishment of religion or, for that matter, the championing of secularism that would naturally form the opposing pole of such a dimension.

Despite these lines of fit between the two levels of evidence, the second level does show somewhat greater heterogeneity that seems to bear on differences in the current salience of the respective cleavages. The relationships between voting preference and language use are clear enough at the first level, but at the second level the differentiation in perceived party positions, while in the "expected direction" is really quite blurred by comparison with the sharp differences in perceptions for the Left-Right and moral-religious issues. The current-issue materials contain other features to underscore this observation. Thus, for example, our tables show that about three respondents out of every four see the taxation and religious instruction issues as matters of partisan controversy, whereas only one out of three hold such a perception on the language question, a disproportionate number of them falling among the bourgeois parties. Moreover, parallel differences emerge in the degree to which people say they themselves have some clear opinion on these various issues. For taxation this figure is 88 percent of party supporters, and for the religious instruction issue it is 87 percent. However, only 72 percent report comparably clear decisions on the language issue.

In short, the implication seems to be that whatever the historical situation may have been, the socio-cultural cleavage surrounding the language problem has declined in salience. Whether the decline is permanent or temporary cannot be said; but the evidence at least suggests that its absolute vitality has been fading, and relative to economic and religious axes of dispute it has probably become quite feeble.

4. The Structure of Perceived Party Locations in 1965

It is in many ways a routine and familiar procedure to show, as we have done in the preceding section, that people who identify with a particular party in a system differ in characteristic demographic ways from those who identify with other parties, and that there are appropriate differences among the various groups of partisans in their positions on current issues.

However, one of the purposes of the 1965 study was to probe beyond these conventional results to what might be thought of as the "deeper party structure" as it may exist in the minds of the Norwegian electorate. The obvious way to do this is to ascertain not only which single party the voter feels closest to, but also the way in which he views each of the other parties available to him in the system. In a system as well endowed with serious political parties as in the case in Norway, it would be time-consuming indeed to collect detailed information concerning the voter's image of and reactions to all seven parties. But at the very least it is possible to ask him quite economically to say something of his fuller preference hierarchy over all of the parties, so that we know what party he would find most congenial after his own, second most congenial, etc., on to the party or parties from which he feels most completely distant, and for which he would vote under no conceivable

circumstances. Such partial preference orders were indeed collected for respondents
in the 1965 survey.18

A number of practical purposes for such an investment could be cited. At the simplest level, for example, not all parties provide slates in every election for all constituencies. While the failure to contest a district usually means that the party lacks local support there, in most such cases there would nonetheless be some remnant of partisans left without their preferred party. Knowledge of their second preference gives a surrogate measure of their choice among available alternatives.

However, our main interest in these preference hierarchies is more analytic, and relates to broad questions of system shape and system change. Where the latter is concerned, let us imagine that the Labor Party were to suffer some major national disgrace of sufficient proportions to alienate at least some part of its normally large clientele in the mass public. What other parties would tend to profit from such an event, and in what relative proportions? Mere knowledge that one or another percent of the electorate identifies with the Labor Party would cast no light on this problem. But valid data on second preferences should indeed provide some worthwhile indications.

Or again, we can turn the matter around and see such second-preference data as highly useful for the development of party strategy. If a substantial proportion of the massive base of Labor support had moral-religious concerns, as witnessed by many second choices of the Christian People's Party, whereas few other voters in the smaller middle parties made such a second choice, it might suggest to strategists of the Christian People's Party that they could strengthen their vote support most readily by adjusting their party program in a socialist direction, so that they could attract Labor voters oriented primarily to economic cleavages but having moralreligious concerns in the background as well.

For any of these problems involving potential change in the system, we must first have a view of the relative distances perceived between the various parties in the system at a given point in time, or the momentary structure of perceived party locations. It is our purpose in this section to develop such a picture, at least as a kind of 'average view' drawn across all of the electorate at once.

The Diagnostic Value of Second Preferences: An Illustration. It is useful at the outset to assure ourselves that within various groups of partisans who are themselves homogeneous in their first preferred party, statements of party preferred as a second alternative do seem to be meaningful. We cannot, with the 1965 study taken alone, show that these preferences are correlated with lines of party change. But we can see whether within a homogeneous set of partisans, second choices of party follow the lines of cleavage we have already established.

The Labor Party is most useful as an illustration, since it provides a large number of cases for subdivision. Moreover, there is a reasonable degree of heterogeneity among Labor supporters as to their second preference, suggesting quite different 'wings' of the Party. Nearly 85 % of Labor's rank-and-file were able to choose a single party as a second preference.19 Within this set of people, the second choices are distributed as follows:

4 % Communist Party
37 % Socialist People's Party
33 % Liberal Party
8 % Christian People's Party
7 % Agrarian Party
11 % Conservative Party

Thus we see that nearly three-quarters of Laborites would prefer either the Liberals or the Socialist People's Party after Labor itself, and this fact conforms with the kinds of judgments that outside observers might have made, i.e., that the Liberal Party and the Socialist People's Party were the closest neighbors to Labor in the party system. However, it is of no small interest that the two main wings of the Labor Party are as nearly balanced in size as they appear to be. If Labor had suffered some great disgrace at this time, it is likely that its alienated membership might have moved off in both directions, although the judgment in any specific case would have to take account of the possibility that the terms of the disgrace might be more keenly felt by one wing than another. Similarly, it is interesting to note that many more Laborites would sooner move to the more remote bourgeois parties than to the Communists, just 'one socialist party away'. This is a strong hint that the several parties are not seen by Laborites as equally spaced along a single left-right dimension. The sheer distance out to where the Communist Party is must seem especially forbidding.

Can it be shown that these second preferences still reflect our cleavage lines in meaningful ways? One obvious test is to see whether these different apparent wings of the Labor Party, as defined by second preferences, show the kinds of differences in issue position that our earlier analyses would lead us to expect. On the issue of progressive taxation, for example, Table VI clearly shows that indeed, Labor Party supporters who favor a steeply progressive rate of taxation are far more likely to prefer some party to the left of Labor as a second choice than is true for people who have no opinion or who are unfavorable. It is interesting that more than half of Laborites who indicate 'no opinion' on the issue, prefer some middle party as a second choice. The proportion mentioning the Conservative Party in second place is nearly three times as large for Laborites who dislike sharply progressive rates as for

those who favor them. The contrasts in the table are very striking, particularly when it is remembered that we are dealing here with a secondary party choice among a group of Laborites who, in any normal tabulations, would be lumped together as politically "homogeneous."

Somewhat similar results can be found in connection with the other cleavage indicators. In Table VII, for example, Laborites favoring more religious instruction in the schools are far more inclined to prefer the Christian People's Party or some other middle party as their second choice than is true for Laborites who think two hours for religious instruction are enough. People of the latter conviction show second preferences which lean more strongly toward the left than toward the center. Again in Table VIII, where respondents are classified according to their stand on language fusion, similar lines appear. The Liberal Party is most often cited as a second choice by Laborites who favor a fusion into samnorsk, whereas the Conservative mentions are relatively frequent among Laborites who prefer a free language development. While the patterns are quite weak in this instance, the differences remain unmistakable.

All told, it is clear that the second party choices among people who are all fundamentally Labor supporters are meaningful not only in distinguishing two wings on the left-right dimension, but in terms of the other classic cleavage lines as well. It would not do to reify the notion of 'wings', because from cleavage to cleavage, intra-Labor differences may show quite a different composition. Moreover, Laborites

differ in their homogeneity of position on various of the issues. Of those who take a stand, for example, 73 percent favor the present taxation system, and 30 percent are in favor of a language fusion, whereas 49 percent support extended religious instruction. Thus the camps within Labor are far from monolithic blocs of fixed size and composition. But the important point for our current purposes is the simple fact that second party choices continue to express familiar cleavages.

The Clarity of the Preference Hierarchies. The preceding analyses assure us that as we probe more deeply into preference hierarchies, we are not picking up haphazard mentions given by voters who never think beyond their own party allegiance to establish coherent preferences across a wider set of parties. Nevertheless, it could scarcely be imagined that all members of the electorate in Norway have constructed in their mind's eye an unambiguous rank location from first to seventh to include all seven of the competing parties. At the very minimum, we might expect that numerous people with strong likes and dislikes for certain parties might nonetheless consider the remaining parties as lumped together without discrimination, or 'tied in rank', somewhere in between. Indeed, this is why we merely required people to order the three parties they preferred and the two that they liked least: we were willing to leave the two unmentioned parties (logically, the fourth and fifth preferences) as ties in rank. Some voters would undoubtedly have been able and even eager to state a preference between the two unnamed parties in fourth and fifth rank. However, this slight gain in information hardly seemed worth the strain placed on the many other voters who would have substantial 'indifference regions'.

Even multiple ties in rank pose no serious problem for our analysis, since such ties can be assigned a meaningful level between ranks one and seven. Thus for a party identifier who sees only one competing party as a bete noire, the remaining five parties can be considered as tied at rank four in a large indifference region.

Somewhat more of a problem, however, is the likelihood that for numerous voters some of the smaller and less visible parties were simply not taken into account in stating relative preferences, either because their salience was so limited or because the respondent was actually unaware of what they stood for or even of their very existence. Where the salience of a party is so feeble, it is not preposterous to let it fall into the central indifference region (around rank four). Nevertheless, since it is

the smaller parties that are most frequently forgotten, there is some tendency for
their average ranks to be less extreme than might have been the case if we had
refreshed the respondents' memories about them.20

We have elsewhere discussed some of the perceptual 'principles' that are likely to account for whether party preference hierachies are full and clearly-discriminated, or are blurred by large zones of indifference or unrecognized parties.21 In general, of course, this kind of visual acuity within the party firmament depends mainly on involvement and attentiveness to politics, which in turn tend to increase with education.22 Some of these effects are demonstrated if we consider how many specific parties the various respondents single out for mention in trying to fill out top three and bottom two preferences, as in Table IX. Overall about 13 percent of the sample - largely nonvoters - fail to indicate a single preference, even 'own party'. And although the relationship is not striking the expected progression as a function of education does appear.

We would expect fully as much gain in clarity of party perceptions to result from strong motivational commitment to party competition, of the kind that is well registered in our measure of strength of party identification.²³ The relationship is indeed stronger, with 56 percent of 'strong' identifiers naming at least, four parties, as opposed to only 11 percent among the 'apoliticals'. Moreover, the fullness of party rankings shows an independent association with both partisanship and education when the two are varied simultaneously (Table X).

In most of the subsequent operations that we have carried out with our preference order material, we have deleted those cases in which fewer than two parties were given some rank, and otherwise have treated unmentioned parties as tied at their logical place in the order. The deletions, which constitute some 32 percent of the total sample, are particularly unfortunate, but they do reflect a political reality which lies beyond our control.

The Full Shape of the Party Space: 1965. While we have used the second-preference materials to develop some impression of how party configurations appear to various Labor supporters, it is interesting to put all of our preference-rank materials together at the same time for supporters of all parties, in order to draw some inference as to the relative distances perceived to separate the seven parties in the Norwegian system as of 1965, from the point of view of a hypothetical 'average' voter.

One way to proceed toward this goal is to calculate the average of rankings which

voters of a given party, like Labor, accord to each of the other six parties. If a particular party seems to be a close neighbor, such as the Socialist People's Party is for Labor, then such an average should take a low value (if the second preferred party is scored as one rank away, the third two ranks away, and the least preferred is scored as six ranks away). On the other hand, a party sharply disliked by most Labor members would turn out to have a rather high average rank. We shall interpret these average values as an index of the degree of perceived distance which supporters of each party consider to lie between themselves and the other party.

It is important to keep in mind that we have no way of beginning with absolute' distances. That is, a voter who prefers Party A slightly over Party B, Party B slightly over Party G, Party G slightly over Party D, but finds Party E a complete abomination that he would place at the greatest possible remove from his own position, will ideally give the same simple ranking input as another voter who may discriminate in the same order but who would sense by far the greatest distance to lie between Party A and Barty B, with only tiny discriminations between G, D and E.²⁴ Thus there is an important sense in which consensuality of high or low ranks accorded by members of a particular party is being interpreted as a straightforward function of some more underlying distance. One implication of this use of ranks as input is the fact that the sum of 'distances' to each of the other parties in the system from any given party is a constant. Therefore if members of a party felt that some other party was gravitating closer toward them over time, the perception could only be expressed by concomitant increases in average ranks for one or more other parties, even though there might have been no perception of absolute change in the other cases.

Since there are seven Norwegian parties with supporters generating such rankings, we can construct a 'similarity-dissimilarity' or distance matrix of seven rows and columns, summarizing all of the average rankings, as in Table XI. In this table we have placed the parties in order from left to right as they would have been judged to be arrayed in 1965, but it is of course no coincidence that this ordering of rows and columns tends to draw the small values (neighboring parties) toward the main diagonal, while pushing the large values out toward the upper right and lower left corners of the matrix. These contours in the table already assure us that the basic left-right cleavage accounts for a lion's share of the particular inter-party distances registered in the matrix.

Table XI departs from a more conventional distance matrix, such as a table for distance between major cities, in that it has asymmetric entries: the distance between Communists and the Socialist People's Party as perceived by the Communists is only a value of 1.85; the same 'distance' perceived from the other end (by the Socialist People's adherents) is a much greater value of 3.22. How can we account for two such different estimates of what should be the same distance?

There are in fact several answers. First, of course, we should remember that all of the entries in the matrix are no more than sample estimates. The subsample of Communists is very small, and while the Socialist People's adherents are not numerous, they outweigh the Communists by more than a factor of three. Thus the

estimate of 3.22 is considerably more stable than the estimate of 1.85, as an
expression of the distance between the two parties.

However, it is likely that the difference in estimates is large enough to exceed any reasonable sampling variability. Moreover, there is no need to be blind to the fact that we are not talking about physical distances, but psychological ones, and it is nothing unusual to discover that such estimates may vary according to one's vantage point. This is particularly true where the Communists are concerned. It will be noted that every party sees itself as farther away from the Communists than the Communists see in return. A similar configuration emerged for comparable matrices in both France and Finland.2s To some degree, the row of values perceived by the Communists could not look much different overall because it must sum to a constant value: even if Communist partisans felt themselves to be a world away from the rest of the Norwegian parties, they would still be obliged to pick some party in each of the adjacent ranks. However, the contrast in values across the first row and down the first column is not a pure artifact. The members of non-Communist parties are in broad consensus as to the remoteness of the Communists, whereas the Communists show less consensus as to what the more distant parties are. The latter fact in itself might attest to a good deal of psychological distance, if the Communists themselves felt so remote from the rest of the parties that the latter looked somewhat alike.

Although there are a number of other discrepant pairs of entries in the portion of the table not involving the Communists that are worth speculative comment, the general fit between corresponding estimates of the 'same' distances is reasonably good, and we shall proceed to our consideration of the table as a whole. Numerous techniques of multi-dimensional scaling exist which will operate on a matrix like Table XI, reducing it to the minimal number of dimensions that would account for all 30 entries within any specific margin of error. Since we are interested in the primary dimensions or cleavage lines that differentiate the several parties, such reductions are of great value to us.

We have submitted the matrix of Table XI to a number of these competing routines. Solutions are obtained which look roughly similar across techniques and display statistics measuring the degree of fit (or limitation of error in accounting for the entries) which are moderately good.26 However, we arrive at still more unequivocal results if we convert the table into a symmetric matrix by assuming that interparty distances are truly identical, from whichever vantage point they may be seen. We accomplish this by forming averages of corresponding pairs of estimates, weighted to take account of the differing number of viewers contributing to each estimate in the pair.

In Figure 4 we present the two-dimensional solution of such a symmetric which is clearly best of those examined, in terms of statistical goodness of fit to the data.27 It shares with all of the solutions obtained a number of broad characteristics. Parties of the left typically fall on a rough line with substantial distances between them. This line 'points' toward the four bourgeois parties, which fall in a cluster or diamond to the right. Among the bourgeois parties it is typically the Christian People's Party that moves farthest away and in effect forces the solution into a second dimension. The same pressure 'outward' into a second dimension can be seen in the values down the Christian People's Party column of Table XI, which tend to run higher than the left-right location of the party taken alone would warrant. In most solutions, also, it is the agrarian 'Center' Party which is its closets neighbor.

Substantively, of course, Figure 4 looks very much as we would expect on the basis of earlier discussions. The left-right axis does indeed dominate the solution, and this for several reasons. It is, as we have seen, the most sharply bipolar cleavage, and is probably somewhat reinforced from more than one source. Moreover, it currently enjoys the highest relative salience of any of the conflicts. The second dimension depicted in Figure 4 is unquestionably the moral-religious one, which we earlier found to have high salience in 1965 as well. We also noted that current positioning of the parties in the conflict was rather unipolar. In point of fact, Figure 4 suggests that both the Communists and the Socialist People's Party are antagonists to the Christian People's Party along this dimension. However, both of these parties are tiny: about 95 % of the electorate are involved in the other five parties. If we look at those five parties alone, we see that there is indeed the rather clear 'T' shape that one might expect if a second, unipolar dimension were crosscutting a first, bipolar one. Only the agrarian Center Party mars the T, and it lies where it should, to the right and relatively religious.

The two main dimensions of Figure 4 are ones which we most expected to find, but they do not comment very directly on at least two of the other cleavages historically important in the system: the urban-rural conflict and the language or socio-cultural cleavage. It is tempting to seek signs of these other cleavages by solving our matrix from Table XI for third and fourth dimensions. We have in fact examined these higher-order dimensions, and the agrarian Center tends to take an extreme position on the third dimension, hinting at some urban-rural conflict line. However, the character of the third dimension turns out to be highly unstable from solution to solution, and even though the Agrarians usually do figure at an extreme, the rest of the locations are not readily interpretable. In one sense this is not surprising: we have seen that Figure 4 provides an excellent fit for the data, and what remains after the effects of the first two dimensions are taken out must be largely measurement error.

Nevertheless, there is another possible way to proceed. We have argued earlier that some of the less salient cleavages date from earlier periods and have had their main effects in keeping the bourgeois parties from joining forces with one another against the socialist parties. If we set the left aside entirely and consider how the members of each bourgeois party regards the other three parties, we neutralize much of the powerful organizing force of the left-right axis, and thereby permit other dimensions dividing the bourgeois parties to come into view. The one pitfall in such reasoning is that as we reduce the number of objects to be located down to a number as small as four, it is easy to get a very good fit in a dimension or two, and higher-order dimensions become meaningless.

We have however looked at such a solution within the bourgeois parties taken alone. As might be expected, the religious cleavage dominates the picture. The Christian People's Party is relatively remote at one extreme, and the Conservatives are the clearest opposition at the other end of the axis. However, the solution is more than just a recapitulation of the bourgeois quadrant of Figure 4. The Liberal Party falls neatly between the Conservatives and the Christian People's Party on

the same axis, and the Center (Agrarian) Party protrudes out into a second dimension, forming again a "T. It seems reasonable to interpret this second dimension in the bourgeois case as some reflection of an urban-rural cleavage, although the only party truly distinctive in this regard is the Agrarian Center.

In the next section, more detailed analyses of the full preference orders will bring some of those residual bourgeois cleavages into clearer view. In the current format, however, we cannot safety press the data any farther. We have seen that the leftright dimension clearly dominates the space of party competition in the Norwegian system, as perceived by the public, and that as of 1965, at least, the moral-religious conflict added a second potent and cross-cutting line of division. There is reason to believe that the other cleavage lines are salient chiefly within the bourgeois camp, and of these, the urban-rural may well be the next most important.

However, it is necessary to keep in mind that the map of party locations in Figure 4 has all of the liabilities as well as the assets of an overall average. It should be recalled that we have averaged ranks within parties, and then averaged rank distances across differing party perspectives, in order to arrive at such a picture. There is nothing wrong with any of those operations, and they constitute a valid means of depicting a central tendency in viewpoint as calculated overall system members at once. Indeed, they give us the most general 'cognitive map' of the parties that we can devise for the total Norwegian electorate in 1965, and it is significant that this map accords well with the summary impressions of the system held by outside observers.

Nevertheless, any average conceals variation, and we have no assurance that any significant proportion of Norwegians would actually perceive the party configuration in just this way. Political perception is very much a matter of vantage point, and we have already noted that the party system looks rather different from the perspective of different 'home parties'. Naturally, partisanship is only one variable shaping the individual's view of party competition. In the ensuing section we shall explore some of the other broad factors that lead to differentiations in the way the party system is seen.

5. Variations in the Perceived Configurations of the Parties

In the last section we found that more specific parties were likely to be ranked among people more strongly identified with one or another party in the system, presumably because they feel more stake in party competition, and give it more attention. This increased clarity at the individual level need not mean that there is greater clarity in perceived party configurations more collectively taken, for clarity in this collective sense must rest on consensus between individuals within the same parties and a higher degree of 'fit' between perceptions from the vantage point of different parties. Nevertheless, it would not be entirely surprising to find greater clarity in this collective sense among the strongly identified than among the inattentive, and indeed differences of this kind are not at all hard to demonstrate.

If we consider two matrices like Table XI, one based on the set of strongly identified partisans and the other based on non-identifiers who gave rank orders, there are quite significant differences in general properties between the two. Thr matrix for strong identifiers shows sharp contrast, with very low or favorable values for neighboring parties down the main diagonal and very high or distant ones toward the corners. The matrix for non-identifiers, while it also shows some valley toward the middle, is generally much flatter. It is clear within parties, at least, there is much greater consensus as to the surrounding party landscape for the strongly committed.

Across parties, as well, there is more agreement among strong partisans as to the basic configuration of parties making up the system. In discussing Table XI we noted that there were some differences in pairs of estimates as to the 'distance' between two parties, as seen from one end or the other. One measure of cross-party 'fit' in perceptions is the correlation between corresponding estimates above and below the main diagonal, a correlation considerably flawed by the case of the Communists, treated earlier. For the strong identifier matrix, the correlation (rho) is .76, but it falls to .58 among non-identifiers. The most stringent test of fit across parties, however, takes not only pairs of observations, but their full pattern at once, as in the multidimensional scaling routine. When that routine is applied to the two matrices, the 'stress' coefficient measuring goodness of fit is considerably better for the identifier matrix than for the non-identifiers. At every level, then, from individual clarity through party consensus to system-wide consensus, cognitive maps of the party system are much more blurred among non-identifiers than among strongly committed partisans.

Apart from focus and detail, are there more substantive differences in perception between the identified and the unidentified? There is only one of any magnitude, but it has considerable interest. Although it is true in general that the distance matrix for the non-identified is cast in flatter relief, there is one exception. Extraordinarily high values (distant ranks) appear for Laborites and adherents of the Socialist People's Party in the average ranks accorded the Communist Party, whereas strongly identified members of these parties see the Communists much more as a neighboring party. However dimly the unidentified of the non-Communist left may see the party system, they are remarkably agreed that the Communist Party is undesirable.

This major discrepancy in perception leads in turn to one basic difference in the general character of spatial solution between the strongly identified and the unidentified. For the strongly identified, the three socialist parties are drawn much more closely together than is the case in Figure 4, and this cluster is more clearly set apart from the bourgeois parties than even Labor is in that Figure. In other words, for strong partisans taken alone, the main cleavage in the system is that which divides socialist parties from the burgeois camp: the perceived distance between any bourgeois party and any Socialist party is greater than the distance between any parties within each cluster. For the unidentified, while some gap is present between the burgeois parties on one hand and Labor or the Socialist People's Party on the

other, the Communist Party is set off in a position quite remote from either
grouping.

It is this kind of variation in perception that underlies the average mapping of Figure 4, and which we wish to explore in some detail over the course of this section. Basically, our interest lies in perceptual differences across groupings of the Norwegian electorate defined in sociological or demographic terms, such as region, urban-rural residence, or age cohorts. First, however, it is useful to see how the shape of perceived party distances changes as a direct function of the individual's position on some of the secondary cleavage issues with which the system is confronted.

One way of proceeding is to examine differences in the contours of distance matrices defined on two opposing groups, much as we have just done for strong identifiers and non-identifiers. We may, for example, take distance matrices for the set of people who drink and think alcohol restrictions too severe, as opposed to abstainers concerned about temperance. If we make a new matrix of numerical differences in rank between the two first ones, we get a clear picture of contrasts in preference order structure associated with the different attitudes toward temperance. Along any given row, the gains and losses in average rank must cancel out to zero, since if one party is moved up a rank, another party must be demoted by a rank as well. However, averaging differences down the columns of such a derived matrix gives a picture of overall 'gains' or 'losses' in ranks encountered by the respective parties between the two attitude conditions.

We present such a summary for three cleavage items in Figure 5. The first two of the items relate to the moral-religious cleavage, and it is not surprising that, as an average across parties, the Christian People's Party is awarded a gain of almost one full rank among involved teetotalers by contrast with drinkers restive about temperance. In the interchange, it is primarily the Conservative Party which loses, although there is considerable loss to Labor as well. Alternatively, we could say with equal legitimacy that the Conservatives gain a rank, and Labor more than half a rank, among people annoyed by temperance regulations by contrast with concerned abstainers, while the Christian People's Party suffers nearly a rank loss.

It is important to keep in mind that the effects which we register in Figure 5 are merely residual ones which linger on in clear form after some other central features of the relevant cleavage is controlled away. Where the temperance question is concerned, for example, we have seen earlier that the temperance composition of the clientele of various parties differs radically (see Figure 2, above). However, all differences of this sort are effectively controlled from view in Figure 5, since we treat the several parties as units, as though they had equivalent composition on the issue. Similarly, the upward adjustment in rank for the Christian People's Party among teetotalists comes exclusively from partisans of other parties than the Christian People. As a matter of fact, Christian People's adherents - the strongest group of moralists in the system - make no contribution whatever to the temperance or religious involvement portions of Figure 5, since no drinking or religiously inactive Christian People's members exist to provide differences in average rank. More generally, however, it is true that own party preference makes no contribution to the

Figure, however strongly that choice may have been influenced by the cleavage
line involved: the Figure deals exclusively with persons' regard for other parties
deeper in their own preference hierarchy.

What all of this means in spatial terms is simply that strong temperance people, for example, perceptually draw the Christian People's Party closer to their own position; while those not interested in temperance perceptually push it off to a greater distance. The spatial solution for the temperance or non-temperance extreme groups taken alone are somewhat degenerate, with only a loose fit achieved. In both cases, however, a considerable semblance of left-right divisions remains. For the temperance enthusiasts, the Christian People's Party is drawn into a rather tightly clustered 'in-group' with the other two middle parties (Liberals, Agrarians), as might be expected from the joint gains of these three parties in the temperance case in Figure 5. This in-group cluster is widely separated from a semi-circle of non-_temperance, 'out-group' parties: the Conservatives to one side, and the three socialist parties in right-left order circling around to the other. For people annoyed by temperance restrictions, on the other hand, the Christian People's Party is somewhat more isolated at an extreme, with the Agrarian Center as nearest neighbor, and the rest of the array in more familiar left-right order.

Figure 5 suggests that modifications of rank orders as a function of differences in religious activity follow more or less the same pattern, although the gains in rank among religious activists occur more exclusively for the Christian People's Party, and compensation tend to be distributed more equally across the four parties to left and right of the middle, although trade-offs with the Conservatives remain relatively prominent.

The fact that coherent adjustments are made in the deeper structure of preference orders in response to the language conflict where it is salient (bottom portion, Figure 5) is of particular interest in view of the fact we were unable to isolate any reliable signs of that cleavage either in our spatial solution for the general electorate, or in that for the bourgeois parties taken alone. We see that these adjustments are indeed weaker than is the case for either of the earlier items which contribute to the moral-religious cleavage. However, the patterns here are not trivial and provide an unmistakable fit with the history of the language problem. The main cleavage is within the bourgeois parties, with the three middle parties (and especially the Liberals) receiving favorable adjustments in ranking among people preferring the samnorsk fusion, and the Conservatives standing as the bastion of riksmål resistance to such reform. The only surprising portion of the pattern is the gain in rank by the Labor Party on the riksmål side, since the socialist parties generally are rather uninvolved in the language conflict. However, closer examination of the full difference matrix suggests that such gains are rather passive ones, coming more or less by default. The samnorsk coloration of all three middle parties sweeps out a considerable portion of the left-right spectrum for people of a riksmål preference. Some of them who downgrade these parties on language grounds have no choice but to upgrade Labor in the process, assuming they are unlikely to want to spread their favors still farther to the left.28

The language problem is somewhat related to the territorial cleavage between center and periphery, in addition to its focus for the socio-cultural conflict. We do not have other direct items of a current-issue type as diagnostic of the territorial cleavage. However, it is possible to produce distance matrices for regional divisions of the sample, as well as for contrasting urban and rural subgroups. In view of the different fates that various of the Norwegian parties have encountered in different areas, it would be surprising if it could not be shown that there is an interesting variation in the perceived shape of party competition along these lines as well.

Regional Perspectives on Party Locations. Beyond the lines of left-right cleavage, there are dramatic contrasts from region to region in the distribution of the three 'counter-cultures' of the language (landsmål) movement, the temperance movement, and the fundamentalist movement.²⁹ These disparities are clear in Table XII, which shows that the strongholds of these counter-cultures lie in the South and West, and most particularly in the rural areas of this region. The sheer similarity in the South and West of the proportions of the electorate who are teetotalists, landsmål supporters and religiously active suggests that the same persons tend to have all three of these characteristics. In fact, a high correlation does exist on an individual level between these postures, and the correlation is stronger in the South and West than it is in the other regions. Therefore while the language conflict is quite conceptually distinct from the moral-religious tension, the two are sufficiently interwoven in some areas like the South and West not to emerge from our analyses as independent dimensions.

The regional contrasts of Table XII leave a strong imprint on aggregate vote distributions. One useful way of summarizing such aggregate votes in order to highlight the local intensity of the classic left-right cleavage is to deal in terms of the total vote proportions given to the three socialist parties on the left extreme and the Conservative Party to the right. This kind of a left-right 'polarization' score ran as high as 80 percent in 1965 for the Eastern region, which is the most industrialized

part of the country. In the Middle and the North the same score is 72 percent, but in the South and West it is only 56 percent, suggesting the strong intrusion upon the left-right cleavage of the other conflict lines. Moreover, in all regions the polarization score is higher in urban than in rural areas. It reaches a maximum of 86 percent in the urban East, and dips in the rural areas of the South and West to a low of 49 percent.

All of these kinds of information attest to the likelihood that lines of cleavage have differential salience by region, a fact which in turn means that the shape of the perceived party system should show regional variation, with the classic leftright axis at its strongest in the urbanized East, and weakest in the South and West. Party distance matrices calculated within these two contrasting regions separately, do show a number of differences, particularly when the urban portion of the sample in the East is paired against the rural portion of the sample from the South and West.

The limits on the rank-order metric means that it is hard to show that the chasm between parties of the left and right are in an absolute sense wider in the East than in the South and West. Indeed, for urban areas in the two regions, the average rank distance between the Conservatives at one end of the left-right spectrum and the socialist parties at the other does not differ greatly. However, when the rural areas of the South and West are examined separately, there are clear signs that the longest distance - that between Conservatives and Communists - is foreshortened by attention to other lines of conflict:

Generally speaking, in rural areas the socialist parties are seen as looking more
similar to one another than they are taken to be in urban areas where the left-right
axis on which they are arrayed is more salient.

As might be expected, the three middle parties contribute the most interesting regional differences in perceived location. These differences occur both in the interrelationships of the three parties, as well as in the way the three parties are located relative to their neighbors in the system to the right and left.

Since the three middle parties represent various combinations of traditional values (rural fundamentalism and a distinctive Norwegian language) that tend to co-occur most clearly in the South and West, it is not surprising that these parties are less clearly differentiated in that region than they are in the East:

As we see, the pairwise party distances are about half an average rank smaller in
the South and West than in the East. They are somewhat smaller still in rural
areas of the South and West taken separately.

Judged on the basis of independent information, we would expect the individual parties to have rather similar substantive connotations or images across regions, with one major exception. This exception is the Liberal Party. During the last decades of the nineteenth century, Liberals in the East and especially Oslo began to drift in a secular direction, thereby splitting increasingly away from the South/West wing of the Liberal Party which remained deeply concerned about traditional religious and moral values. In Tröndelag and the North the character of the Liberal Party was more mixed, but if anything remained rather traditional.

These differences between the two wings of the Liberal Party were clearly reflected in the 1965 survey.³⁰ Thus in the South and West, Liberal voters show religious activity and temperance concern beyond the national average, whereas in the East Liberals fall below the average for both items. In fact, in the Oslofjord area, Liberals are quite similar to the Conservatives in their indifference to temperance concerns. Such regional variation in the moral-religious outlook of Liberals fits party history very well and is therefore not surprising. Less familiar has been the discovery that the Liberal Party in the South and West has a clientele which is quite undistinctive occupationally, although slightly overrepresenting the middle class. In the East, however, the Liberal Party is overwhelmingly supported by voters of upper middle class backgrounds. Thus the Liberal Party shows a series of major differences between the two regions, and there is only one major issue on which its partisans do not show marked regional variation. This is the language issue: in both regions Liberal voters are more likely to be landsmål users than are the voters of any other party. These inherited national values are one element which helps to keep the party unified.

The perceived locations of the Liberal Party in the East as opposed to the South/West tend to follow these patterns very clearly in our distance matrices. In the East the Conservative Party turns out to be seen as the closest neighbor to the Liberals (average rank distance of 2.38), followed by Labor (2.79). In the South/ West, however, the Christian People's Party and the Agrarians are the closest neighbors (2.38 and 2.51 units distant, respectively), although Labor and the Conservatives are not far behind. The perceived distance between the Liberal Party and the Christian People's Party is two-thirds of an average rank greater in the East than it is in the South/West.

All told, the rank order data conform with a picture in which the basic left-right cleavage is entirely dominant in the East: parties are differentiated along a fairly unidimensional continuum. Of course the same left-right organizing dimension is perceived in the South/West as well; but political perceptions in this region tend to have a more multidimensional character. Both the moral-religious an the sociocultural dimensions appear to intrude forcefully here, although they are so highly intercorrelated in the area that it would be hard to isolate the independent significance of each.

It is naturally true that distance matrices calculated for other demographicallydefined subgroupings in the Norwegian electorate tend to show their characteristic differences as well. Thus, for example, a matrix for workers shows the socialist parties drawn close together and the bourgeois parties pushed to relatively great psychological distances, whereas the parallel matrix for higher middle class people would show reverse effects. However, differences of this type are of less significance than those discussed between regions, because it appears that both workers and middle class are perceiving the same general configuration of parties, only from different ends of the spectrum. In the regional case, the differences go beyond questions of perspective over the same fixed landscape; the shape of the landscape itself is different.

We presume that knowledge of such difference in the shape of the party landscape itself, going beyond mere questions of perspective, is an important addition to any understanding of potential for change. However, the true fit between perceived party distance and actual change over time has remained quite speculative. The empirical mapping of party cleavages into multidimensional spaces as perceived by the common voter is a sufficiently new enterprise that there are few examples available for various countries even at a single point in time, much less for a succession of points over time. However, we now can go at least one step further.

6. Change in Perceived Party Locations by 1969

The Norwegian program of electoral studies is the first to provide any longitudinal measurements, since preference orders were extracted from the electorate in exactly the same way in connection with the 1969 election survey. In fact, the 1965 and 1969 surveys constitute a panel study because the same respondents were interviewed at both elections. The 1969 data are sufficiently recent that they have not as yet been subjected to more detailed dimensional analysis. However, we have been able to look at the most general solution, parallel to Figure 4 for the situation in 1965, in order to see what mix of stability and change occurs.

Ideally, of course, we would prefer not to find a great deal of change, for if the underlying preference orders were extremely unstable either in response to immediate events or because of biases in measurement, then a measurement at any single point in time near an election would have only limited interest: perhaps a

month earlier or a month later the configuration would have looked different. Furthermore, anything we know about party positions, cleavages and voter preferences would suggest a high degree of stability. There is, of course, drift over time in all of these terms which is of supreme political significance. But if our preference orders were highly unstable in a four-year period, it would suggest that they were failing to reflect what we wish.

On the other hand, we would hope that measurements over time would indeed reflect whatever limited changes one might suppose on other grounds to be taking place. If such measurements were fixed in all particulars over time, especially periods in which party configurations seemed to be evolving in one direction or another, the approach would not be very useful.

Examination of the 1969 data seems to fulfill these hopes on both sides. Table XIII shows a distance matrix derived from the 1969 preference orders, which can be directly compared to the 1965 data in Table XI. It is clear that the basic stabilities are very strong. It we convert this table to the symmetric (or 'triangular') matrix used as input for the multidimensional scaling routines, and compare the size of the entries with those for corresponding cells in the 1965 symmetric matrix, we find a rank-order correlation (rho) of just under .97.

On the other hand, any comparison of Table XIII and Table XI indicates that changes are captured which seem too large and patterned to be considered mere sampling error. Perhaps the most efficient comparison comes from the spatial solution for the 1969 distance matrix, as shown in Figure 6. Although the 'stress' or goodnes of fit of the general 1965 solution (Figure 4) was considered to be excellent, that for 1969 is even better.³¹ Therefore we can be assured that we are looking at highly stable solutions in both cases.

In most regards, the two 'maps' approach identity. There is, however, one prime change: three of the four bourgeois parties appear to have 'imploded' toward a common locus in 1969, by comparison with 1965. It we return to Table XIII, we may conclude that the algorithm somewhat overestimates the completeness of this convergence, since something more than minimal average distance values continue

to separate these three parties in Table XIII. Nonetheless, it is clear that there has been some overall reduction in interparty distances within this triad of bourgeois parties between 1965 and 1969, both from examination of the mutual estimates, as well as from examination of how other parties view these three.

If we consider the relationships between the Norwegian parties at the governmental level between 1965 and 1969, this convergence in voter perceptions of the three burgeois parties is not at all remarkable. For it was during the 1965-69 period that the bourgeois parties combined to operate a coalition government, an entirely new experience in peacetime Norwegian politics. After the 1969 election, substantial tensions arose between the coalition partners, but during the period to which our data refer, cooperation was the dominant pattern of the coalition. The governmeent was able to find compromise solutions to several problems on its agenda, and it made a considerable effort to present itself to the public as a harmonious team.

Nonetheless, the participating bourgeois parties did seek to maintain their individual profiles, and it was clear to any alert observer that old antagonisms between the burgeois parties were never far removed from the surface. Manifest conflicts arose during this period in two main areas. In the first instance, a distinguished leader of the Christian People's Party, Mr. Kjell Bondevik, had become minister of Church and Cultural Affairs. On a number of issues he skillfully promoted policies quite congruent with the platform peculiar to his own Christian Party. In

addition to predictable opposition from the socialist parties, there were also rumblings
of discontent from within the coalition. In particular, the secular (Eastern)
wing of the Liberal Party made clear its criticism of Mr. Bondevik's programs.

In addition, certain controversies began to develop within the coalition following the old commodity market lines of cleavage between urban and rural economic interests. This controversy, articulated mainly as in days of yore by the Conservatives on one hand and the Agrarians on the other, arose in several connections, such as a relevant tax reform and the negotiations toward a Nordic customs union.

Against this backdrop of events, the modest changes between the overall spatial solutions for 1965 and 1969 (Figures 4 and 6) make a good deal of sense. The locations of the three socialist parties have passed practically unchanged, but there has been little reason to expect change in that sector. Three of the bourgeois parties - the Liberals, Agrarians and Conservatives - have come much closer together in public perceptions, a matter which implies some success in their efforts to convey an impression of agreement. They have been obliged to defend the same government and the same policies, and their convergence in the public mind over this period seems very reasonable.

However, the Christian People's Party has not participated in this convergence. Indeed, they appear in Figure 6 to hold if anything a more pronounced position on the moral-religious axis, while the Liberals and Agrarians are accorded a more neutral position. Furthermore, the Liberals appear to have made a move toward the right in order to converge on the Conservatives and Agrarians. This cluster of change is not particularly surprising. They may well reflect voter reactions to the controversies between the Christian People's Party minister of Cultural Affairs and some prominent Liberal members of the Storting. From other data it is known that one of the main changes in voting between 1965 and 1969 involved Liberal Party losses concentrated in the South/West, and preliminary analyses have suggested that most of the defectors from the Liberals have moved over to the Christian People's Party.32

These shifts have made inroads on the proportion of Liberals in the South/West who have strong religious proclivities. The percent of Liberal voters classified as 'high' on religious activity fell from 42 percent in 1965 to a mere 19 percent in 1969. The defections of fundamentalist Liberals to what is now their more natural home in the Christian People's Party also changed the socioeconomic composition of the Liberal Party in the South/West. It left a remaining clientele which was more middle class than it had been before: the proportion of Liberal voters who reported some higher education (at least 3J years beyond elementary school) increaseed from 30 percent in 1965 to 44 percent in 1969. Such changes in a secular and higher-status direction among people ranking the Liberal Party as their first choice almost necessarily implies corresponding changes in the position of the party along both the moral-religious and left-right axes, in the directions actually observed.

There is only one important respect in which the 1965-69 changes in the bourgeois
camp do not fit expectations: there is no sign that antagonisms along the ur-10»

ban-rural cleavage have produced any greater distance between the Conservatives and the Agrarians. In fact, the analysis suggests a tendency in the opposite direction. When a separate multidimensional analysis was applied to the four bourgeois parties with the 1965 data, the urban-rural cleavage emerged clearly as the second most important line of conflict within the bourgeois camp. A similar analysis of the 1969 data did not produce a corresponding result, despite intervening controversies within the coalition reflecting this cleavage. Apparently these particular controversies were not strident enough to capture the public eye or offset the general message of teamwork and harmony between coalition partners.

7. Conclusions

We have been concerned in this paper with the problems of change in political cleavage systems. Due to the dynamic forces of political development, new cleavages emerge from time to time, and some of them produce one or more new parties if the state of the system is conducive. On the other hand, a decline in the salience of a given cleavage may have other substantial effects upon the party system.

We have tried to show that the Norwegian party system is based upon five cleavages that accumulated over time: the territorial division, the socio-cultural opposition, the moral-religious cleavage, and economic conflicts in the labor market and the commodity market. Our analyses have been addressed to the ways in which each one of these conflict lines finds expression in current electoral politics in Norway. In fact, where popular voting is concerned, substantial correlations attest to the existense of these five cleavages. However, further analysis suggest that these cleavages vary greatly in their salience, and some probably now leave no more than geological traces. For some of the dimensions, we have been able to relate partisanship to individual perception of party positions on critical issues. This work clearly demonstrates that people are able to relate themselves to the left-right (labormarket) dimension, as well as to the religious cleavage. But the picture is far less clear where the language problem (socio-cultural conflict) is concerned, and people now tend to see the issue as nonpartisan. Once a basic cleavage, it has now lost much of its salience.

Our remaining analyses of party preference hierarchies in the electorate generally serve to support these conclusions. The individual's choice of second most preferred party has been shown to have high cleavage significance, and manipulation of the full preference orders over all seven parties permits the establishment of matrices of perceived average distances between all of the political parties. Such matrices can then be subjected to further multidimensional analyses.

The main findings from this work can be briefly summarized:

(1) The left-right and the religious-moral conflicts seem to emerge as the most
salient cleavages in current Norwegian politics.

(2) The socio-cultural cleavage is not isolated as an independent dimension in the spatial solutions. Partly this may due to the fact that position on this dimension (i.e. meassured by language preference) is strongly correlated with position on the moral cleavage (measured by temperance and fundamentalism), and the two dimensions relate to the parties in an identical way. In other words, some of the effects of the socio-cultural cleavage may be blurred by the relationship to the religious-moral conflict. However, the fact that the preference ordering data give results similar to the issue material leads us to believe that the socio-cultural conflict has lostits potency.

(3) Territorial conflicts still exist in the sense that preference orderings data give substantially different structural solutions in different regions. To some extent at least these differences are artifacts because of heterogeneity in the territorial distribution of moral, religious and cultural values.

(4) The conflict in the commodity market is less clear than we had expected. However, it appeared as a third dimension explaining the divisions between the bourgeois parties at the 1965 election. Again this conflict may be blurred by its interrelations with other salient dimensions, e.g. the moral-religious and the left-right dimensions.

(5) A comparison of preference orderings in 1965 and 1969 shows that the cleavage structure has remained remarkably stable over this four year period. However, some minor changes have occurred which reflect political events on the government level. The fact that issues of a moral-religious character have been articulated within the government coalition accounts for the changes in locations of non-socialist parties along the religious axis. Similar changes should be expected along an urban-rural axis, since issues related to this dimension have stirred up a fair amount of controversy within the coalition. But in fact, urbanrural differences are not visible at all in the spatial solutions of the data.

We may speculate why it is that religious differences on the government level are reflected in the perceptions of the voters while urban-rural differences do not come out this way, despite the fact that the latter differences have probably been of a more serious character. The most likely explanation is that controversies involving religious and moral values are easily brought to the attention of the public. They attract great interest and tend to stir up debate. Conflicts concerning economic differences are more complicated and less visible to the public. These are conflicts on an elite level. Perhaps we would have been able to trace the urban-rural conflict if we analyzed the preference orderings separately for the strong party identifiers. We have seen that the clarity of preference orderings increases with increasing identification with some party. If we were able to measure preference orderings on a truly elite level we would probably capture a cleavage structure which is sharper and more consistent with conflict patterns prevalent among political decision-makers. Indeed it would be a challenging task for future research to bridge the gap in cleavage structures between elites and the rank-and-file.

Since the 1969 election a heated debate has taken place in Norwegian politics

concerning membership in the European Common Market. In terms of our cleavages, the urban-rural and the territorial conflicts have become most salient, and the conflict along the left-right axis has changed in character. Within the coalition government the issue has created open antagonisms, particularly between the Conservatives and the Agrarians. And internal conflicts created by the EEC-question were obviously a major reason when the coalition resigned in March 1971. Present conflict lines seem to deviate completely from the normal pattern, and the same thing happened in 1961-63 when the question of Norway becoming a member of EEC was first discussed.³³ An alliance between rural interests (i.e. interests of farming and fisheries), particularly in the peripheral parts of the country on the one hand and on the other hand the radicals in the cities, particularly leftist socialists, constitute the main opposition against membership in the Common Market. Most in favor of membership are forces with open communication lines with the political and economic centers of Europe. The Conservative Party with the employers organizations belong in this group together with the bulk of the Labor Party leaders and large parts of the trade union movement.

We have to return to the 1880's to find a cleavage structure similar to the present
one.

It is interesting to notice how relationships to other nations may effect the domestic cleavage structure. In the 1890's when the conflict over the Union with Sweden was the dominant issue in Norwegian politics, a rather strange alliance occurred between the Conservatives at the centre and the fundamentalists in the periphery. After World War I the Russian revolution had a radicalizing effect upon the Labor party which split up in three parties. Again during the Cold War Norway's membership in the Western Military Alliance caused a great deal of antagonism within the Labor party, and finally led to the establishment of the Socialist People's party in 1961. Whether or not Norway becomes a member of the Common Market, it is very likely that the present debate has already had a durable impact upon the cleavage system. It will be interesting to see whether future data bear out such a prediction.

---

NOTES

1 Anthony Downs, An Economic Theory of Democracy. New York: Harper and Brothers, 1957.

2 Davis, 0., and Hinich, M., 'A Mathematical Model of Policy Formation in a Democratic Society', in Mathematical Applications in Political Science, 11. J. Bernd, (Ed.), Southern Methodist University, 1966.

8 Converse, Philip E., 'The Problem of Party Distances in Models of Voting Change', in M. Kent Jennings and H. Zeigler, (Eds.), The Electoral Process. Englewood Cliffs, New Jersey: Prentice-Hall, 1966, pp. 175-207.

4 See H. Valen and S. Rokkan, 'The Norwegian Program of Electoral Research', in Scandinavian Political Studies, Vol. 11. Helsinki, The Academic Bookstore, 1968. A nationwide sample of 1751 voters, drawn by probability methods, were interviewed before the 1965

NOTES

1 Anthony Downs, An Economic Theory of Democracy. New York: Harper and Brothers, 1957.

2 Davis, 0., and Hinich, M., 'A Mathematical Model of Policy Formation in a Democratic Society', in Mathematical Applications in Political Science, 11. J. Bernd, (Ed.), Southern Methodist University, 1966.

8 Converse, Philip E., 'The Problem of Party Distances in Models of Voting Change', in M. Kent Jennings and H. Zeigler, (Eds.), The Electoral Process. Englewood Cliffs, New Jersey: Prentice-Hall, 1966, pp. 175-207.

4 See H. Valen and S. Rokkan, 'The Norwegian Program of Electoral Research', in Scandinavian Political Studies, Vol. 11. Helsinki, The Academic Bookstore, 1968. A nationwide sample of 1751 voters, drawn by probability methods, were interviewed before the 1965

---

election, and 1623 of them were re-interviewed after the election. The same sample was interviewed before and after the 1969 election. Most of the analyses reported in the present paper are based on 1965 data.

5 See S. Rokkan and H. Valen, 'Regional Contrasts in Norwegian Polities', in E. Allardt and Y. Littunen, (eds.) Cleavages, Ideologies and Party Systems. Helsinki, Westermarck Society, 1964 (Reprinted in E. Allardt and S. Rokkan, (eds.), Mass Politics. New York, The Free Press, 1970); S. Rokkan 'Geography, Religion, and Social Class' in S. M. Lipset and S. Rokkan (eds.), Party Systems and Voter Alignments. New York, The Free Press, 1967; and H. Valen 'De politiske partiene' (political parties) in O. Jonsrud and T. Sandberg, (eds.), Politisk ABC, Oslo. Gyldendal, 1969.

6 See particularly F. A. Hermens, Democracy or Anarchy? University of Notre Dame Press, 1938; and M. Duverger, Political Parties, New York: Wiley, 1954.

7 The capitalized abbreviations for the major cleavages are used in our summary historical Table I, p. 113.

8 The geographical realignments associated with the territorial cleavage are discussed in detail in S. Rokkan and H. Valen, 'Regional Contrasts in Norwegian Polities', op. cit.

9 The role of proportional representation in permitting the establishment of the Agrarian Party has been openly admitted by the first Secretary-General of the Party, Hans Holten. S enter partiets inf ormas jonsblad, Oslo 1958, no. 3, p. 5.

10 This has been convincingly demonstrated by G. Øidne, 'Litt om motsetninga mellom Austlandet og Vestlandet' (about the contrasts between East and West), Syn og Segn, 63 (3), 1957.

11 For a much more detailed account of the historical evolution of these cleavages see Stein Rokkan, 'Geography, Religion and Social Class', op. cit.

12 See S. Groennings, Cooperation among Norway's Non-Socialist Political Parties. Ph. D. Dissertation. Stanford University, 1962. Microfilm 62-5480, University Mrirofilms, Ann Arbor, Michigan.

13 See H. Valen, 'De politiske Partiene', op. cit.

14 The respondents were asked: "Our tax laws are such that people with high incomes must pay a considerably larger part of their incomes in taxes than people with low incomes. Many people think that the pressure of taxation is far too heavy on people with above-average incomes, while others think it is only fair that those who are better off must pay such high taxes. Is this a question on which you have tried to reach a decision? (IF YES) What do you think, are our existing taxation laws fair or do you feel that this taxation hits people with somewhat higher incomes rather hard?"

15 The specific items were, first: "Do you consider yourself a temperance person or do you happen to take a drink from time to time?" Those choosing the temperance description were asked about their relative interest in the temperance issue. People who would drink were asked "As you know", the authorities in this country have put into effect many measures regulating the use of alcohol. What do you think of these rules: are they too strict or too lenient?" The temperance index of Figure 3 combines these items on a dimension from relatively strong interest in temperance to those nonteetotalers displeased at the strictness of present regulations.

16 The index of religious activity is basedon the number of religious meetings and services attended during the preeceding month; the number of religious programs followed on radio or television; and membership in religious organizations.

17 The questions were: "Then there is the question of religious instruction in the schools. Many people think that religious instruction should be limited to the number of hours laid down in the 'Normal Plan for Schools' (about two hours a week), while others maintain that the individual communes should be allowed to introduce a greater number of hours than is laid down in the 'Normal Plan'. Is this a question on which you have tried to reach a decision?" (IF YES) "What is your attitude toward religious instruction? Do you think two hours a week is sufficient, or would you like the individual communes to be allowed to introduce a somewhat greater number of hours?" And then: "Which party or parties do you think would adopt the attitude closest to your own when this topic comes up in the Storting?"

18 Vote intention (in the pre-election interview) was recorded as first preference. In order to measure the second and third choice the respondents were asked: "Not thinking of (party given for vote intention) which party do you like best?" "And which one after this do you

election, and 1623 of them were re-interviewed after the election. The same sample was interviewed before and after the 1969 election. Most of the analyses reported in the present paper are based on 1965 data.

5 See S. Rokkan and H. Valen, 'Regional Contrasts in Norwegian Polities', in E. Allardt and Y. Littunen, (eds.) Cleavages, Ideologies and Party Systems. Helsinki, Westermarck Society, 1964 (Reprinted in E. Allardt and S. Rokkan, (eds.), Mass Politics. New York, The Free Press, 1970); S. Rokkan 'Geography, Religion, and Social Class' in S. M. Lipset and S. Rokkan (eds.), Party Systems and Voter Alignments. New York, The Free Press, 1967; and H. Valen 'De politiske partiene' (political parties) in O. Jonsrud and T. Sandberg, (eds.), Politisk ABC, Oslo. Gyldendal, 1969.

6 See particularly F. A. Hermens, Democracy or Anarchy? University of Notre Dame Press, 1938; and M. Duverger, Political Parties, New York: Wiley, 1954.

7 The capitalized abbreviations for the major cleavages are used in our summary historical Table I, p. 113.

8 The geographical realignments associated with the territorial cleavage are discussed in detail in S. Rokkan and H. Valen, 'Regional Contrasts in Norwegian Polities', op. cit.

9 The role of proportional representation in permitting the establishment of the Agrarian Party has been openly admitted by the first Secretary-General of the Party, Hans Holten. S enter partiets inf ormas jonsblad, Oslo 1958, no. 3, p. 5.

10 This has been convincingly demonstrated by G. Øidne, 'Litt om motsetninga mellom Austlandet og Vestlandet' (about the contrasts between East and West), Syn og Segn, 63 (3), 1957.

11 For a much more detailed account of the historical evolution of these cleavages see Stein Rokkan, 'Geography, Religion and Social Class', op. cit.

12 See S. Groennings, Cooperation among Norway's Non-Socialist Political Parties. Ph. D. Dissertation. Stanford University, 1962. Microfilm 62-5480, University Mrirofilms, Ann Arbor, Michigan.

13 See H. Valen, 'De politiske Partiene', op. cit.

14 The respondents were asked: "Our tax laws are such that people with high incomes must pay a considerably larger part of their incomes in taxes than people with low incomes. Many people think that the pressure of taxation is far too heavy on people with above-average incomes, while others think it is only fair that those who are better off must pay such high taxes. Is this a question on which you have tried to reach a decision? (IF YES) What do you think, are our existing taxation laws fair or do you feel that this taxation hits people with somewhat higher incomes rather hard?"

15 The specific items were, first: "Do you consider yourself a temperance person or do you happen to take a drink from time to time?" Those choosing the temperance description were asked about their relative interest in the temperance issue. People who would drink were asked "As you know", the authorities in this country have put into effect many measures regulating the use of alcohol. What do you think of these rules: are they too strict or too lenient?" The temperance index of Figure 3 combines these items on a dimension from relatively strong interest in temperance to those nonteetotalers displeased at the strictness of present regulations.

16 The index of religious activity is basedon the number of religious meetings and services attended during the preeceding month; the number of religious programs followed on radio or television; and membership in religious organizations.

17 The questions were: "Then there is the question of religious instruction in the schools. Many people think that religious instruction should be limited to the number of hours laid down in the 'Normal Plan for Schools' (about two hours a week), while others maintain that the individual communes should be allowed to introduce a greater number of hours than is laid down in the 'Normal Plan'. Is this a question on which you have tried to reach a decision?" (IF YES) "What is your attitude toward religious instruction? Do you think two hours a week is sufficient, or would you like the individual communes to be allowed to introduce a somewhat greater number of hours?" And then: "Which party or parties do you think would adopt the attitude closest to your own when this topic comes up in the Storting?"

18 Vote intention (in the pre-election interview) was recorded as first preference. In order to measure the second and third choice the respondents were asked: "Not thinking of (party given for vote intention) which party do you like best?" "And which one after this do you

---

like best?" The lowest rankings were measured by the following questions: "Which party do you like least, so little in fact that you could not vote for it under any circumstances?" "Is there any other party you dislike very much?"

19 The conceptual significance of incomplete preference orders will be discussed shortly. In this instance, an additional 6—7 percent of Laborites gave the other two socialist parties or some bourgeois parties as tied for second preference. Only 9 percent were unable to provide any second choice.

20 It is likely that the Communist Party, however, has exceptional visibility given its small size in Norway, due both to local history and its unusual relevance for the international scene.

21 See Converse, P. E., 'The Problem of Party Distances in Models of Voting Change', op. cit.

22 A convincing demonstration that information about politics as well as general awareness of political phenomena are positively associated with level of education is provided by Willy Martinussen, Politisk Oppmerksomhet og Sosial Bakgrunn (Political Awareness and Social Background), in Tidsskrift for Samfunnsforskning (9), 1968: 253—274. See also Martinussen's article in the present volume of Scand. Pol. Studies.

23 Respondents were asked "Would you say that you generally consider yourself a Conservative, a Laborite, a Liberal, etc., or don't you consider yourself attached to any party?" Those indicating some attachment were then asked: "Do you consider yourself to be a strongly convinced supporter of your party, or aren't you very strongly convinced?" People reporting strong party identification constituted some 35 percent of the total sample, and weak identifiers amounted to about 33 percent; 28 percent indicated that they are non-partisans or "independents," whereas 4 percent are classified as "a-politicals".

24 In the course of more recent studies, not yet ready for report, efforts have been made to permit differentiations in distance between the perceived party locations.

25

26 Since all of these multidimensional techniques involve lengthy iterations from a starting point which has random elements, solutions vary somewhat from trial to trial even with the same technique. However, the more clear and simple the dimensionality underlying such a matrix, the more stable the solution. We shall only deal here with solutions that are highly stable.

27 This is a two-dimensional solution from the MDSCAL (multidimensional scaling) computer routine due to J. B. Kruskal. The 'stress', or measure of goodness of fit, for this solution is an excellent figure of .002. See J. B. Kruskal, 'Multi-Dimensional Scaling by Optimizing Goodness of Fit to a Nonmetric Hypothesis', Psychometrika 29 (1964), pp. 1—27; and Nonmetric Multidimensional Scaling: A Numerical Method', Psychometrika 29 (1964), pp. 115-130.

28 Something of the same sort may be happening in connection with the temperance issue in Figure 5. It may be noted that the bars to either side of the vertical do not balance off perfectly in 'weight' in Figure 5 despite the overall constraint that an increase in the ranking of one party must result in the demotion of another. This constraint is of course present, and operates powerfully on the patterns in these graphs. The balance is not perfect, however, because parties have been taken as units in the averaging, without regard for differences in size.

29 See S. Rokkan and H. Contrasts . . .' op. cit.

Valen, 'Regional 30 See H. Valen, 'De politiske partiene', op. cit.

31 Where zero would reflect a perfect fit, the stress coefficient for 1965 was .002, and for 1969, .001.

32 See S. Rokkan and H. Valen: 'The Election to the Norwegian Storting in September, 1969', Scandinavian Political Studies, vol. 5, 1970.

33 The changing conflict lines during the 1961-63 EEC-debate has been dealt with by S. Rokkan and H. Valen: 'Regional Contrasts . . .' op. cit.

like best?" The lowest rankings were measured by the following questions: "Which party do you like least, so little in fact that you could not vote for it under any circumstances?" "Is there any other party you dislike very much?"

19 The conceptual significance of incomplete preference orders will be discussed shortly. In this instance, an additional 6—7 percent of Laborites gave the other two socialist parties or some bourgeois parties as tied for second preference. Only 9 percent were unable to provide any second choice.

20 It is likely that the Communist Party, however, has exceptional visibility given its small size in Norway, due both to local history and its unusual relevance for the international scene.

21 See Converse, P. E., 'The Problem of Party Distances in Models of Voting Change', op. cit.

22 A convincing demonstration that information about politics as well as general awareness of political phenomena are positively associated with level of education is provided by Willy Martinussen, Politisk Oppmerksomhet og Sosial Bakgrunn (Political Awareness and Social Background), in Tidsskrift for Samfunnsforskning (9), 1968: 253—274. See also Martinussen's article in the present volume of Scand. Pol. Studies.

23 Respondents were asked "Would you say that you generally consider yourself a Conservative, a Laborite, a Liberal, etc., or don't you consider yourself attached to any party?" Those indicating some attachment were then asked: "Do you consider yourself to be a strongly convinced supporter of your party, or aren't you very strongly convinced?" People reporting strong party identification constituted some 35 percent of the total sample, and weak identifiers amounted to about 33 percent; 28 percent indicated that they are non-partisans or "independents," whereas 4 percent are classified as "a-politicals".

24 In the course of more recent studies, not yet ready for report, efforts have been made to permit differentiations in distance between the perceived party locations.

25

26 Since all of these multidimensional techniques involve lengthy iterations from a starting point which has random elements, solutions vary somewhat from trial to trial even with the same technique. However, the more clear and simple the dimensionality underlying such a matrix, the more stable the solution. We shall only deal here with solutions that are highly stable.

27 This is a two-dimensional solution from the MDSCAL (multidimensional scaling) computer routine due to J. B. Kruskal. The 'stress', or measure of goodness of fit, for this solution is an excellent figure of .002. See J. B. Kruskal, 'Multi-Dimensional Scaling by Optimizing Goodness of Fit to a Nonmetric Hypothesis', Psychometrika 29 (1964), pp. 1—27; and Nonmetric Multidimensional Scaling: A Numerical Method', Psychometrika 29 (1964), pp. 115-130.

28 Something of the same sort may be happening in connection with the temperance issue in Figure 5. It may be noted that the bars to either side of the vertical do not balance off perfectly in 'weight' in Figure 5 despite the overall constraint that an increase in the ranking of one party must result in the demotion of another. This constraint is of course present, and operates powerfully on the patterns in these graphs. The balance is not perfect, however, because parties have been taken as units in the averaging, without regard for differences in size.

29 See S. Rokkan and H. Contrasts . . .' op. cit.

Valen, 'Regional 30 See H. Valen, 'De politiske partiene', op. cit.

31 Where zero would reflect a perfect fit, the stress coefficient for 1965 was .002, and for 1969, .001.

32 See S. Rokkan and H. Valen: 'The Election to the Norwegian Storting in September, 1969', Scandinavian Political Studies, vol. 5, 1970.

33 The changing conflict lines during the 1961-63 EEC-debate has been dealt with by S. Rokkan and H. Valen: 'Regional Contrasts . . .' op. cit.