Winfried Kurth

A quantitative approach contributing to the monitoring of public mood, based on the evaluation of cartoons from newspapers

(May 2000)



Since 1995, a special interest group inside the German Society for Psychohistorical Research (German branch of the International Psychohistorical Association) is collecting political cartoons from a fixed sample of German and Swiss newspapers in order to identify and to analyze trends and patterns in nation-wide public moods and group fantasies. This project was motivated by the wish to apply the method of group fantasy analysis developed by Lloyd deMause (1979, 1982, 1984, 1990) to the German situation and to provide a continuous monitoring of mood trends which could possibly help to identify "critical situations" and turning points of the group process in an early phase. The huge number of pictures which did soon accumulate, as well as the seemingly disparate and ambiguous contents which excluded an easy identification of clear trends, led to some frustration during the early meetings of the interest group, until a quantitative approach based upon counting certain simple motifs was developed by the author. This method was originally meant only as a preliminary device to facilitate the identification of trends and cyclic patterns in the material, and particularly the choice of motifs to be monitored was considered as provisional and as a first attempt. However, the obtained results and the long data series obtained with this tool so far have had the consequence of a certain "conservation" of the method. In fact, adding new motifs or changing the way of evaluation would have meant to re-evaluate thousands of cartoons collected since 1995 under the new premises in order to generate an equally consistent and equally long data series - a task which no member of the special interest group dared to undertake. But this pragmatic conservatism does not mean that we consider the method optimal. Several modifications were already proposed and will be discussed later in this paper.

Furthermore, it has to be emphasized that the quantitative analysis of mass data is only a part of our investigation of group fantasies, and it is not the most deep-going one. Qualitative methods are indispensable to give interpretations and meaning to the identified patterns. Concerning the role of statistics and its relation to other methods in psychohistory, I agree with Ralph Frenken: "A combination of methods... is not a narrowing to positivistic arguments, but a widening of allowed interpretational approaches by using several converging scientific ways of arguing" (Frenken 2000).

This paper concentrates on the quantitative method. It describes therefore the process of analysis of public moods and group fantasies only partially. The method was developed at German and Swiss material, but a short-period study of U.S. newsmagazines (3 months in 1996, see last part of this paper and Kurth 1998b) gave promising evidence that it can be applied to other nations, too. Some of the results will be mentioned in the end, but concentrating only on the special ascpect of quality assessment of the method; for a more detailed analysis and discussion of group fantasies derived from this approach and from qualitative interpretations the reader is referred to other papers (Kurth 1997, 1998a, 1999, 2000a,b).

The investigated pictures

In a first phase of investigation (10 months, from April 1995 to January 1996), 13 newspapers and newsmagazines were taken into consideration (cf. Kurth 1996): Frankfurter Allgemeine Zeitung (FAZ), Frankfurter Rundschau (FR), Göttinger Tageblatt (GT), Handelsblatt (HB), Neues Deutschland (ND), Der Spiegel, Süddeutsche Zeitung (SZ), Tagesanzeiger (TA), Die Tageszeitung (taz), Der Tagesspiegel (TS), Die Welt, Weltwoche (WW), Die Zeit. Two of these newspapers, TA and WW, are edited in Zurich (Switzerland), the rest in Germany. Different regions and different political directions are covered, but no attempt was made to ensure some kind of representativeness for Germany as a whole, since our group had not the resources to investigate an even larger sample. FAZ and Welt are conservative newspapers, HB is targeted to business and finance, FR has positions close to the Social Democrats, the taz was originally a self-organized project of the Alternative movement and is still close to the Green party, the ND is associated with the former communists of East Germany, Zeit and Spiegel are liberal newsmagazines. Zeit, Spiegel and WW are issued weekly, whereas the other newspapers have daily issues (HB: only from Monday to Friday, TS and Welt: every day including Sunday, all others: from Monday to Saturday). From February 1996 on, we restricted the sample of newspapers from 13 to 8. This restriction was done on the basis of a correlation analysis which will be discussed below. For certain periods, other newspapers were also considered, e.g. the Westdeutsche Allgemeine Zeitung (WAZ), but these were not included in the final quantitative evaluation.

Normally, all political cartoons from each issue of each newspaper were taken into consideration - with the exception of the Spiegel, for which only the front page was considered. Not considered were comic strips, advertisements, reproductions of pieces of art, fotos (except they were part of the Spiegel front page) and drawings serving purely as naturalistic portraits, without political or ironical intent. Until March 2000, a total number of 17,716 cartoons have been investigated according to the schematic method which is explained in the next paragraph.

The key motifs

To manage the large number of pictures, we tried to develop a categorization or identification of key motifs enabling a rapid evaluation of cartoons which should be as unambiguous and transparent as possible. On the other hand, the method should at least partially open a possibility to get access to unconscious contents which were supposed to be present in the public imagery according to deMause (1982) and other psychohistorians (Atlas 1992, Dervin 1994, see also Link 1990). After considering the pictures collected so far, the participants of the second meeting of the special interest group undertook a brainstorming to produce a list of 57 motifs which had been found to appear in the material. Example keywords from this list are: depression, pressure, erection, idealization, farewell, rubbish, victim, pregnancy, funnel, vagina dentata, water, butchering. Such an unsystematic list of 57 keywords still seemed to be ineffective for a comprehensive evaluation of the material. Finally, the author reduced the list to 7 key motifs. These are graphical motifs in the pictures, which can be characterized shortly by the following key words, but which need to be described in a more detailed way to ensure proper recognition (such a description will be given below):

         D      dominating, unendangered person,
         T      rising trend, ascent, flight,
         M      the strong D-Mark,

         E      narrowness, encirclement, confinement, being tied,
         Z      cracking, crumbling, disintegration (German: Zerfall),
         A      slipping, falling, crash, abyss,
         V      being devoured.

These motifs are not to be understood as a categorization in which all cartoons must fit. In fact, a significant part of the investigated cartoons (around 50 %) remains without "motif match". On the other hand, it happens sometimes that several of the motifs appear together in one and the same cartoon.

For the final quantitative evaluation, the 7 motifs were grouped into two classes: motifs of strength (D, T, M) and motifs of threat (E, Z, A, V). The motif M ("strong D-Mark") was relatively rare from the beginning and has practically disappeared from the German newspapers during the last years (the introduction of the Euro being one of the reasons). Hence it will not be taken into account in future studies, and we arrive at 2 motifs of strength and 4 motifs of threat. These will be explained in more detail in the next paragraphs. We present typical examples as well as "border cases" and sources of error in the process of identification of the motifs. In order to make the results as reproducible as possible and to ensure compatibility with our German data obtained so far, it is recommended that these explanations should be studied carefully by anyone who wants to apply the same method to new material.

D     Dominating, unendangered person

One person (his whole body, or only his head or face) dominates the picture. Other persons, if they appear at all, are distinctly subordinated to the main person (Figure 1) or placed in the background. A typical example for motif D is the "Karl May" front page of the Spiegel from May 1st, 1995 (Fig. 2). Here, the dominance of the central person is underlined by additional attributes (weapons, clothes) and by posture and bearing. However, this is not in any case necessary for motif D to be present. Hence in Figure 3, despite of the somewhat ridiculous attributes, the central person (former German minister of finance, Theo Waigel) dominates the picture with his activity.



Fig. 1. ND, September 13, 1996:
Motif D.

Fig 2. Spiegel, May 1, 1995:
Motif D.

Fig. 3. Welt, July 6, 1996:
Motif D.

Fig. 4. GT, April 4, 1995:
Motif D.

Fig. 5. SZ, January 9, 1997:
Motif E.

Fig. 6. ND, December 2, 1996:
Motif D.

The dominating person may also be a "rogue" (Fig. 4). At a deeper level, such a "dominating rogue" may represent parts of the recipient's self, and the transitions to a "positive" (?) identification figure like Karl May (see Fig. 2) are flowing. The only condition which must be fulfilled by the person who dominates the picture is: He must not be endangered. For example, the Milosevic depicted in Figure 5 is - despite of the ignition layer which he still controls - in a situation of encirclement, hence we have motif E here (and not D). In contrast, the Milosevic of Figure 6 is not in danger, but appears as an active juggler. A background knowledge about an eventual political risk threatening the depicted person is not taken into account. The evaluation concentrates very much on the visible, graphic content of the cartoon under consideration. Hence, texts will principally also not be taken into account. For example, the bubble in Figure 7 says "First signs of decay", but the picture tells us the opposite, and we have motif D again.

Fig. 7. ND, Oct. 21, 1996:
Motif D.

Fig. 8. GT, Oct. 24, 1996:
Motif D.

Fig. 9. ND, July 4, 1996:
None of the motifs.



Fig. 10. WAZ, April 16, 1996:
None of the motifs.

Fig. 11. HB, Nov. 19, 1996:
Motif T.

Fig. 12. HB: Motif T.

Confrontations of two persons, or struggles, are generally not consistent with motif D, except the case when one person is distinctly dominating or already the clear victor (Fig. 8). Difficult border cases result when the looser is depicted in detail and can also be seen as an identification figure. In the case of doubt the presence of motif D is to be neglected (Fig. 9). Likewise, D is to be excluded when several strong persons are present in one picture (Fig. 10).

T     Rising trend, ascent, flight

Typical examples for this motif are ascending rockets (Fig. 11), airplanes or balloons. But principally, already the "minimal version" of the rising trend, depicted in Figure 12, is sufficient. Additionally, all visualizations of the positively loaded dream of "flying" belong to this type of motif (e.g. Fig. 13). The only condition is that no danger of crash is manifest in the picture - otherwise we have motif A (see below). Interestingly, border cases between these two motifs are rare, i.e. normally one can decide quite clearly whether a (potentially euphorizing) high flight or a (threatening) situation of abyss is present. Border cases are more likely to occur in situations when the motif of the "rising trend" (in the form of the diagonal of the picture) or the "high flight" are present only in a hidden, subtle form. Figure 14 shows an example. Here we can recognize motif T, although in a rather obscured form. (The author is grateful to Reinhard Merker for bringing this example to his attention.)

Fig. 13. TA, September 6, 1995:
Motif T.

Fig. 14. WAZ, November 18, 1996:
Motif T.

Z     Cracking, crumbling, disintegration

A person, or a building or other object, shows distinct traces of crumbling or decay, manifested normally in cracks, fissures, scraps lying around, missing parts or even in total disintegration. Typical examples are shown in Figs. 15 and 16.


Fig. 15. GT, June 27, 1995:
Motif Z.

Fig. 16. GT, December 7, 1996:
Motif Z.

Frequently, this motif is associated with images of quarrel and chaos, e.g. in Figure 17, an example from the United States, where it manifests itself in the torn banner ("Dole 96"). However, not every quarrel implies motif Z. For instance, in the "mutiny" depicted in Figure 18, some small panes of glass get broken, but the essential parts of the ship and of the "coalition building" are not in danger, so that this confrontation has a distinctly play-acting character (also indicated by the clothes), and the disintegration is not yet serious.



Fig. 17. The New Republic, Nov. 25, 1996: Motif Z.

Fig. 18. WAZ, October 29, 1996:
None of the motifs.


Fig. 19. TS, April 26, 1995:
Motif Z.

Fig. 20. WAZ, September 7, 1996:
Motif Z.

Fig. 21. WAZ, October 26, 1996:
Motif Z.

Fig. 22. SZ, October 21, 1996:
Motif Z.

Normally, for motif Z to be present, the cracking or disintegrating element should play an important role in the picture and should attract the viewer's attention (Figs. 19, 20). The disintegration can also be expressed in a temporal sequence (Figs. 21, 22) or in the form of dissolution or liquefying (Fig. 23).


Fig. 23. HB, September 29, 1995:
Motif Z.

Fig. 24. WAZ, November 29, 1996:
Motif Z.

Fig. 25. GT, April 12, 1996:
None of the motifs.

Fig. 26. TS, March 5, 1996:
Motif Z.

We have border cases when the disintegration is actively perpetrated by a person (Fig. 24, see also Fig. 21) and when the motif of cracking is present only in a weak form or at the margin of the picture (Fig. 25). On the other hand, the mutilation can have extremely brutal forms (Fig. 26).

A       slipping, falling, crash, abyss

A person stands directly at the edge of a dangerous hole or abyss (Figs. 27, 28), hangs perhaps in a precarious manner at the edge (John Major in Fig. 28, former German minister of finance Theo Waigel in Fig. 29) or is already falling down (Clinton in Fig. 30) or just crashed (Fig. 31).


Fig. 27. TS, June 7, 1995:
Motif A.

Fig. 28. TA, July 8, 1995:
Motif A.


Fig. 29. SZ, May 22, 1996:
Motifs A and Z.

Fig. 30. Washington Post Weekly,
August 5-11, 1996: Motif A.

Fig. 31. Welt, November 12, 1996:
Motif A.

Fig. 32. GT, April 8, 1995:
Motif A.

Fig. 33. TS, June 5, 1995:
Motif A.

Fig. 34. TS, May 16, 1995:
Motif A.

Frequently it is some vehicle which approaches the abyss (Fig. 32). Sufficient for motif A is already a distinct slipping, a dangerous act of balance (Fig. 33), or e.g. a possible falling down from horseback (Fig. 34).

When we have a fall into a narrow hole, the borders to the motifs E and V are blurring; in Figure 35 we will still recognize motif A. Motif A should dominate the picture. In contrast, in Figs. 36 and 37 it is relativized by the reappearing person, resp. by the tank, and it becomes more or less marginal. In Figure 38 a crash has taken place, but the main person (former SPD leader Oskar Lafontaine) is obviously not harmed at all, so that we can recognize "only" a kind of immobilization and ties (motif E).

Fig. 35. TS, July 11, 1995:
Motif A.

Fig. 36. WAZ, April 20, 1996:
None of the motifs.

Fig. 37. taz, April 23, 1996:
None of the motifs.

Fig. 38. TS, March 26, 1996:
Motif E.

Fig. 39. WAZ, November 12, 1996:
Motif A.

Fig. 40. SZ, January 29, 1996:
None of the motifs.

The crash or the abyss must indeed be dangerous (Fig. 39); in contrast, in Figure 40 one can recognize more a sporting challenge than an abyss.

Considering the keyword "abyss", one could be tempted to project some sorts of symbolic "abysses" in a transferred, psychic sense into the pictures. However, this is not intended in this quantitative phase of analysis. Like the other motifs, motif A has to be understood in a graphical, representational manner. E.g., neither the "hole in the budget", which is referred to in Fig. 41 in a cautious and indirect way, nor the subject of social decline symbolized by the self-confident beggar in Fig. 42 imply the presence of motif A.

The motif "being threatened by drowning", which is not very seldom, is not counted in an extra class, therefore we subsumed it under motif A (Fig. 43). However, it would probably make sense to differentiate in a more detailed way by counting it separately.

Fig. 41. WAZ, September 5, 1996:
None of the motifs.

Fig. 42. GT, April 13, 1996:
Motif D.

Fig. 43. WAZ, November 2, 1996:
Motif A.

V     Being devoured

Here, the viewer or one or several persons with whom he can identify himself are on the verge of being devoured by a huge mouth or throat (Figs. 44, 45). It is sufficient if this danger is imminent (Fig. 46).

Fig. 44. TS, June 13, 1995:
Motif V.

Fig. 45. FAZ, October 2, 1996:
Motif V.

Fig. 46. WAZ, November 20, 1996:
Motif V.

Fig. 47. Welt, September 14, 1996:
Motif V.

Like in the case of motif D, the identification figure can also be a "rogue" (Fig. 47). However, by far not every oral action, not every eating person implies the presence of motif V (Fig. 48)! V is reserved for the case when an identification figure enters himself a throat - a situation which is felt as particularly threatening; in the view of depth psychology we could assume a connection to the birth trauma. Following this idea further, it would probably be correct to identify this motif also in the case when someone jumps through a hoop (Fig. 49) - but the threatening character of the situation is too weak to see "V" here.

Fig. 48. ND, November 18, 1996:
Motif D.

Fig. 49. HB, September 29, 1995:
None of the motifs.

We have other border cases when no identification figure is visible, or when the devouring monster presents itself as a figure to identify with. E.g., in Figure 50 we have to weigh the pros and cons, and we will probably draw the conclusion that throat and teeth of the lion are depicted in a manner dangerous enough to trigger the fear to be eaten.

Fig. 50. WAZ, September 23, 1996:
Motif V.

Fig. 51. TS, April 28, 1995:
Motif E.

E     narrowness, encirclement, confinement, being tied

This motif is present when one (Figs. 51, 52) or several persons (Fig. 53) are trapped by a rope or in a web, or if they are tied (Fig. 54).

Fig. 52. TS, June 27, 1995:
Motif E.

Fig. 53. TS, January 3, 1996:
Motif E.

Fig. 54. Washington Post Weekly,
August 5-11, 1996: Motif E.

Fig. 55. WAZ, December 5, 1996:
Motif E.


Fig. 56. TS, March 19, 1996:
Motif E.

Fig. 57. WAZ, April 17, 1996:
Motif E.

Sufficient for this motif is e.g. already for a person to be held at a lappet of his suit (Fig. 55). The motif is also present when - like in Fig. 56 - an identification figure (face!) is distinctly enclosed, squeezed or in a situation of narrowness, or when an encirclement or confinement takes place (in Fig. 57 in a rather subtle form). However, it is not sufficient when only a circle is present, without encircled persons, like in Figure 58.

Fig. 58. WAZ, October 4, 1996:
None of the motifs.

Fig. 59. TS, August 5, 1995:
None of the motifs.

Fig. 60. WAZ, November 28, 1996:
None of the motifs.

Fig. 61. WAZ, December 11, 1996:
None of the motifs.

Likewise, we do not have motif E in the following cases: Only a wall is erected (Fig. 59), the encircling attacks are compensated by a distinct robustness of the central figure (Fig. 60), or the object of confinement is not a person but only an unpersonal thing (Fig. 61).

When someone is object of an only slight physical disturbance - like former minister of the exterior, Klaus Kinkel, in Fig. 62 - we have not yet motif E. If we recognized it in case of such subtle disturbances, drastic cases like Fig. 63 would be "devaluated".

Of course, some border cases can occur, particularly regarding the motif of "confrontation" (which was not yet counted in our evaluation scheme). So the dark clouds in Figure 64 are obviously threatening and aggressive enough to cause a "confinement" of the depicted worker; the confrontations in Figs. 65 and 66, however, lack the element of unavoidability, so that we cannot recognize a true trap for the identification figure (Lafontaine in Fig. 65, the German "Michel" (Mike) in Fig. 66). In both cases there remains a possibility of retreat.




Fig. 62. WAZ, December 21, 1996:
None of the motifs.

Fig. 63. Spiegel, December 30, 1996:
Motif E.

Fig. 64. WAZ, September 27, 1996:
Motif E.

Fig. 65. WAZ, December 2, 1996:
None of the motifs.

Fig. 66. WAZ, October 19, 1996:
None of the motifs.

Fig. 67. WAZ, October 1, 1996:
Motif E.

General remarks concerning the identification of motifs

It can happen that several of the 6 basic motifs appear together in the same picture. However, normally it can be taken for granted that motifs of strength and motifs of threat do not occur in the same picture - because this would contradict the definitions. If, for example, the motifs D and E are assumed to appear in the same picture, the person to whom "D" refers is either not really dominating or not unendangered. Practically, it will become necessary to decide with whom the viewer identifies himself. E.g., in Figure 67 empathic identification with the maltreated population is probable - although for persons with hidden sadistic wishes it would also be possible to identify themselves with the perpetrator, former minister of finance, Theo Waigel. But already the structure of the picture gives the victims at least as much place as is given to Waigel, even if they are presented as smaller individuals. Similarly, in Figure 68 the motif Z is most evident (it is additionally accentuated by the black sky), whereas the "dominating" military person (though strengthened by the phallic appearance of the tank at the right margin of the picture) does not yet justify a classification of the whole picture as showing motif D. In Figures 21 and 22, too (see above), destruction dominates the sequence of pictures, hence we don't have motif D. Even Figure 69 does not yet fall under "D": The most suggestive figure to identify with is the tortured German "Michel" (at least for German viewers), although we have here certainly a border case, since Kohl (on the left) is already depicted as (literally) very "dominating". However, in Figure 70 we have Boris Yeltzin really as a "dominating, unendangered person" filling the picture - the identification with the destroyed houses (or better with their invisible inhabitants) will be too weak to put this impression under question.

Fig. 68. TS, April 27, 1995:
Motif Z.

Fig. 69. taz, April 27, 1996:
Motif E.




Fig. 70. SZ, January 27, 1996:
Motif D.

It should have become clear that the motifs are to be interpreted quite literally and in a representational way. The reason is that in the phase of the analysis of mass data an unambiguous, quick recognizability of the motifs, using concrete characteristics of the picture, appeared to be most desirable. In the ideal case we could imagine a computer program which uses automatic image-analysis techniques to identify motifs like the "ascending arrow" in Fig. 12 and performs an automatic classification. But the border cases which we have also presented demonstrate that this is not so easy in practice and that a completely automatic recognition of the motifs must remain an unattainable ideal.

The schematic way of identifying the motifs which is favoured here should, however, not lead to the conclusion that a more subtle (hermeneutic, i.e. working with background knowledge) qualitative analysis of some pictures cannot follow the prestructuring of the mass data. In fact, such a detailed analysis makes often sense and leads to much deeper insights. For our German and Swiss material, we have additionally used headlines from the newspapers to get background knowledge about events co-occurring with the cartoons and to gain understanding of group fantasy development from an analysis of texts. We have made the experience that for a clear identification of group processes (and even more for an understanding of their causation) the schematic counting of the basic motifs is indeed not sufficient. But these quantitative data can play the role of indicators insofar as they refer to points on the time axis where a further investigation can be particularly fruitful - e.g. when certain motifs are unusually frequent.

In our team we have discussed several times to include further motifs or groups of motifs into the quantitative monitoring. Candidates are, e.g., in the category of "strength" the motif "phallus, cannon, rocket" (P; see Fig. 71), which has a long tradition in political cartoons (Kaulbach 1987), or in the category of "threat" the motifs "illness", "poison" or "dirt, smut" (S; see Fig. 72). However, these appear relatively seldom, compared with our basic motifs used so far. Hence we have to decide between a more complicated evaluation key in the mass data monitoring and a potential gain of information which is rather small. Generally, one should be careful with attempts to enlarge the two categories "strength" and "threat" by new motifs, because in an additive evaluation using only these two categories the selectivity of the method could suffer. There is already some evidence (see below) that the category of "threat" is even now somewhat too "noisy" - probably from the rather frequent and emotionally often weak motif "E". It would make sense to test if we get reasonable results already if we restrict the set of motifs instead of enlargening it.

Fig. 71. TS, February 8, 1996:
Proposed motif P.

Fig. 72. TS, June 10, 1995:
Proposed motif S.

Till now, we have not yet included in our evalutation scheme the confrontations and struggles which we have already mentioned several times (Fig. 73) or depictions of more complicated human relations (Fig. 74). Other motifs not yet covered are children or animals. But all these motifs cannot readily be subsumed under the gross categories of "strength" and "threat". A special evaluation would be necessary for them, and this could make sense.

Fig. 73. Focus, June 3, 1996:
Confrontation (none of the motifs).

Fig. 74. Focus, August 5, 1996:
None of the motifs.

We did not explain the 6 basic motifs at this length here to defend them as superior to other possible evaluation schemes, but to enable the reader to achieve a solid knowledge about the method used here and to give him the possibilty to make his own experiments with it. Especially, it was an intention to trigger investigations of cartoons with several raters working in parallel at the same or at different material.

We will now proceed to describe the method of mass evaluation of cartoons, which has not yet come to an end once a motif is identified in a picture. First of all, for each day and for each newspaper the frequency of each motif is counted - most newspapers contain several cartoons in each issue. To inhibit an artificial weekly periodicity with "peaks" of frequency at certain days of the week, we "spread" the values of the weekly appearing magazines over seven days around the date when they appear printed, dividing the frequencies of the motifs by 7. E.g., the weekly Zeit appears on Thursday, and if 14 is the frequency of a certain motif in one issue of this journal, each day from the Monday before till the Sunday after gets a value of 2.

Then for each day the values obtained for the different newspapers are added together. In this summation, we give the titlepages of the newsmagazine Der Spiegel a threefold weight, in order to reflect their larger size and better public visibility.

We then add the frequencies D + T to give a strength index and the frequencies E + Z + A + V to give a threat index. Clearly, this last step can be criticized for several reasons. The motifs are qualitatively different, and we add qualities which are ultimately incomparable if we put them together this way. Furthermore, some of the motifs are possibly emotionally "stronger" than others. However, an arbitrary weighing of certain motifs ("1 times being devoured = 3 abysses" ?) would be even more questionable. The additive compression to two index values was done for the pragmatic reason of reduction of data complexity; a more detailed evaluation of the different motifs remains nevertheless a task for future work.

A "danger index" was already used by Atlas (1992), however, his index was based on a more informal classification of cartoons according to ad-hoc defined thematic groups; the list of themes was different in each month. Consequently, Atlas' index was calculated only monthly.

Our daily calculation of both index values allows for a continuous depiction of the courses along the time axis with the highest possible resolution of one day. However, for a graphical representation at this level of resolution the courses of the raw index values turn out to be too strongly oscillating or "noisy", and furthermore we have an undesired "weekend effect", because most German newspapers do not appear on Sundays (the Handelsblatt not even on Saturdays), which results in a weekly periodicity of "lows" of both index values. To smoothen the curve and to get rid of the artificial weekend low, we use normally centered 5-days sliding averages, i.e. for each day the arithmetic mean of the index value of the day and the index values of the two preceding and the two subsequent days is used in the diagram. To get an overview of long-term trends, sliding averages with a longer time span can also be used, e.g. monthly averages. As an example, Figure 75 shows the long-term evolution of the strength index, using monthly sliding averages.

Fig. 75. The course of the strength index 1995-1998 (sliding 31 day averages).

With the 5-days average, we have still a rather high resolution which allows to relate many of the "highs" and "lows" of the index curves to political or other public events. On the other hand, the "weekend effect" is considerably reduced. The shape of the curves gives some impression of the up and down of public mood. As an example, we show in Figure 76 a diagram of both index curves encompassing a period of 4 months, from May to August 1997. Events in German and international politics and public life are indicated by arrows.

Fig. 76. Strength index (unbroken line) and threat index (broken line), May - August, 1997 (sliding 5 day averages).

We can see already from the short period visible in Fig. 76 that the strength curve and the threat curve are often nearly reciprocal (how it could be expected), but that they are not exact mirror images of each other. In fact, in certain situations we have simultaneous maxima of both curves, indicating unusual tension or splitting. It is important to recognize that we have two independent variables (in the technical sense); hence the approach for mass data evaluation presented here leads to a two-dimensional phase space of public mood. So this approach, despite all the problematic simplifications inherent to it, allows at the level of mass data still more differentiated statements than e.g. a single "danger index" or other univariate evaluation schemes.

However, there are some technical restrictions resulting from the fact that we have used a very preliminary and unstandardized way of "measuring" trends. First of all, there is no rule of normation for the index values which would allow the arbitrary addition or omission of newspapers from our sample. Each newspaper has its own typical number of cartoons per issue, and possibly there is also a certain editorial "style" which can reflect itself in the frequencies of the motifs, so that the averages can change in a significant way if one newspaper is added or omitted. Insofar, it is not so easy to "open" the index tables for free addition of other newspapers evaluated by volunteers (e.g. via internet). This is a good idea in itself, but it would require at least a kind of normation algorithm to prevent the index from biases caused by random fluctuations in the number of evaluated newspapers and of evaluators. Because the frequencies of the motifs are different from newspaper to newspaper (and, furthermore, can change with time), such a normation algorithms is not easily developed.

But at least it is possible to assess with the technique of cross-correlations if and how much a given newspaper coincides in the tendency of its cartoons with all the other newspapers of the sample: The correlation coefficient between the time series obtained from one single newspaper and the time series constructed from the sum of the values from the rest of the sample is calculated. We did this for the first 10 months of our evaluation and used the results to reduce our sample by omitting those newspapers which were not so "typical", i.e. had low correlation coefficients. (The exact numerical results are given in Kurth 1996.) The reduced sample, used since February 1996, consists of GT, HB, Spiegel, SZ, TA, WW, taz and Zeit. Both Swiss newspapers, TA and WW, belonged to those with the best correlation values. The correlations of the values from the smaller sample with those from the larger one were near 70 percent, while the number of cartoons to be evaluated could be reduced by 45 percent.

Assessing the quality of the method

Different questions can be asked regarding our method of monitoring public mood:

What about the intersubjective reliability of the identification of motifs? Will different persons doing the evaluation of the same material independently from each other come to the same results?

How regular, or, conversely: How "random" (without visible rule) are the index curves resulting from the method? This question aims to pattern recognition - we can ask for the inherent complexity and inner structure of time series, ignoring their origin and significance for the moment. A negative result (no apparent regular pattern) would not devaluate our method, since the laws of public mood can be too complicated (and interrelated with external or psychic driving forces) to be manifested in the form of simply recognizable patterns. A positive result, however, would be a confirmation that our method does at least produce something "non-random", whatever the reasons are.

Are there connections between the highs and lows of the index curves and political events (or other public activities)? Are there correlations with other relevant data, e.g. from polls, or with macroeconomic indicators? In contrast to the preceding question, these questions do not concentrate on the inner structure of the time series, but on their relation to a whole-society context. The relation to economic statistics, polls and other data reflecting developments in society is currently a question of research, and at the time we can only say that we found some evidence that such relations exist, at least concerning the stock market and road accidents (see Kurth 1999). But this evidence has to be hardened further, and will be a topic of future publications. Neither will we discuss here the manifold connections to political events which could be identified (cf. Fig. 76); the reader is referred to other papers (Kurth 1996, 1997, 1998a, 1999, 2000a,b).

The question of reproducibility

Our team has undertaken several attempts to assess reliability of the evaluation by letting several members of the team consider the same cartoons independently from each other. In these experiments, where 4 persons were involved, the amount of coincidence in the classification under the categories of strength and threat ranged between 52 and 68 percent (detailed results in Kurth 1997). However, in these experiments the detailed descriptions of the motifs as they are presented in this paper were not yet accessible to the evaluators. They had only short descriptions at hand, causing some misunderstandings. A recent experiment with a fifth person who had studied the German version of this paper is not yet completely evaluated, but brought apparently a much better coincidence as could be seen from the independently derived diagrams which were very similar.

We can conclude that a careful study of the evaluation scheme is a prerequisite for a secure, reliable application of the method. A phase of training for the evaluators would be advantageous to improve the coincidence. However, a total coincidence will never be reached. To ensure intersubjective stability of the method, it would be desirable to have a team of evaluators working independently at the same material. Only those cartoons where they coincide with their judgement would then be included in the quantitative evaluation. Without doubt, this would improve the intersubjective reliability of the method considerably. However, the parallel evaluation by several raters costs much time, and so our team has not been able to realize it for longer periods.

The inner structure of the index curves

With the help of spectral analysis (SAS Institute, Inc., 1992), we were able to identify non-random periodicities in the course of the strength index in shorter periods (see Kurth 1997). For the whole 5-years period, the structure of the strength curve still deviates from an irregular pattern, showing a weak periodicity of approximately monthly duration (32 days). Since we have no magazines in our sample which appear only monthly, this cannot be an artifact. However, the periodic tendency is much weaker than in the case when only one year or two years are considered. Periodicity seems not to be fixed, but the frequency of highs and lows seems itself to evolve with time. Qualitative investigations, using the notion of Stein cycle (see Kurth 1997, 1998a, 2000b), revealed a shortening of cycle durations in the last years.

The threat curve, on the other hand, showed no significant periodicity in the spectral analysis. It seems to be more irregular than the strength curve. This does not mean that it has no significance. However, we did observe that the peaks and lows of the strength curve corresponded much more often to political events than those of the threat curve. The strength index seems to be more specific than the threat index, an observation which is in accordance with the more patterned course of the strength index. Maybe in the definition of the threat index, too many different motifs have been put together, making it unspecific and erratic. However, very high peaks of the threat index are not meaningless and can be put into correspondence with political developments in several cases. Thus one should ignore the smaller fluctuations of this index and concentrate on its extremes.

International comparisons

Can the method be applied to other countries than Germany and Switzerland? We think the motifs are general enough and can be found also in the cartoons which appear in other countries of the West, maybe even in other cultures (Japan, China). Several newspapers should be included in the sample which has to be chosen; just one newspaper is not enough, because it will not reflect the general trend in a sufficiently reliable manner.

The author has done one first attempt to apply the method to material from the U.S.A. in 1996 (see Kurth 1998b). This study covered only three months (August - October) and included the presidential elections. The U.S. material consisted mainly of title pages of weekly appearing magazines, and both indices were therefore calculated only with weekly resolution. (This drawback should be corrected in future studies by the inclusion of a larger number of daily newspapers.) Despite of this restriction, the comparison of the U.S. curves with the German-Swiss curves in the same 3-months period (Fig. 77) shows already some similar tendencies. Several peaks of the German-Swiss strength index are roughly reflected in the zigzag pattern of the U.S. curve, and particularly the threat maxima of both samples do roughly coincide. Beyond this comparison of diagrams, a qualitative analysis revealed also a coincidence of Stein cycles (cf. Stein 1981), i.e. of deep birth-enacting fantasies, in both nations, preceding the date of the presidential elections (see Kurth 1998a,b).

Fig. 77. Strength indices (unbroken line) and threat indices (broken line) from German-Swiss cartoons (lower curves) and from U.S. cartoons (upper curves). Evaluated U.S. journals: Wall Street Journal, Washington Post Weekly, U.S. News & World Report, Time, Newsweek, The New Yorker, The New Republic. The U.S. indices were calculated only weekly (see text).

If newspaper samples from several countries are evaluated according to the same scheme, it would probably be possible to assess a "mood distance" between countries: How much do they coincide, how great are the differences? Is, e.g., Russia further away (in a psychic sense) than Japan?

We should, however, emphasize that the quantitative evaluation can only give a rough measure of the status of the group. It is probably correct to speak of a measurement of tendencies of public mood, but to analyze group fantasies, further (qualitative) evaluations are necessary. Nevertheless, our method allows a prestructuring of mass data and gives hints to moments when public mood undergoes dramatic developments. By plotting strength and threat feelings in diagrams, it allows an easy visualization of basic mood trends. The time resolution is high enough to identify developments taking place at a daily timescale.


This paper is partly a translation of a contribution to one of the annual conferences of the German Society for Psychohistorical Research (German branch of the IPA), see Kurth 1997. The author expresses his gratitude to Florian Galler, Frank Horstmann, Ludwig Janus, Reinhard Merker and Christian Neuse for their help in collecting cartoons and for many fruitful meetings.



Atlas, Jerrold (1992): Was in Deutschland passieren wird. Econ, Düsseldorf.

DeMause, Lloyd (1979): Historical group-fantasies. The Journal of Psychohistory, 7 (1), 1-70.

DeMause, Lloyd (1982): Foundations of Psychohistory. Creative Roots, New York.

DeMause, Lloyd (1984): Reagan's America. Creative Roots, New York.

DeMause, Lloyd (1990): It's time to sacrifice... our children. The Journal of Psychohistory, 18 (2), 135-144.

Dervin, Dan (1994): Group-fantasy during the early phases of the Clinton administration. The Journal of Psychohistory, 22, 171-186.

Frenken, Ralph (2000): Changes in German parent-child relations from the fourteenth to the middle of the seventeenth century. The Journal of Psychohistory, 27 (3), 228-272, 304-307.

Kaulbach, Hans-Martin (1987): Bombe und Kanone in der Karikatur. Jonas Verl., Marburg.

Kurth, Winfried (1996): Gruppenphantasieanalyse für Deutschland 1995/96 - ein Versuch. In: Ludwig Janus (ed.), Psychohistorie, Pubertät und Identität. Conference proceedings, Textstudio Gross, Heidelberg, 85-95.

Kurth, Winfried (1997): Quantitative und qualitative Ergebnisse der Analyse deutscher Gruppenphantasien 1995-96. In: Edmund Hermsen, Ludwig Janus (eds.), Die psychohistorische Dynamik von subkulturellen Bewegungen am Ende des Jahrtausends. Conference proceedings, Textstudio Gross, Heidelberg, 1-52.

Kurth, Winfried (1998a): Analysis of German group fantasies (1995-1996): An empirical approach. Tapestry, 1 (1), 5-24.

Kurth, Winfried (1998b): Die Wiederwahl Clintons als Ergebnis kollektiver Phantasien in der US-amerikanischen Öffentlichkeit. Amerikastudien / American Studies, 43 (2), 257-284.

Kurth, Winfried (1999): Analyse von Gruppenphantasien des Jahres 1997. In: R. Frenken (ed.), Psychohistorie und Biographik. Conference proceedings, Textstudio Gross, Heidelberg, 139-169.

Kurth, Winfried (2000a): The psychological background of Germany's participation in the Kosovo war. The Journal of Psychohistory, 27 (2), 100-123.

Kurth, Winfried (2000b): Das Projekt "Analyse von Gruppenphantasien in Deutschland" - die Jahre 1998 und 1999. In: Ludwig Janus, Winfried Kurth (eds.), Psychohistorie, Gruppenphantasien und Krieg. Mattes Verlag, Heidelberg, 197-243.

Link, Jürgen (1990): Zahlen Kurven Symbole. Zum Anteil der Kollektivsymbolik an normalisierenden Zahlenspielen. Kulturrevolution, 23 (June 1990), 3-9.

SAS Institute Inc. (1992): SAS system documentation, version 6, 4th ed., Cary, NC.

Stein, Howard F. (1981): Trumpets and drums: Some issues in interpretation and methodology in the study of American group fantasy. The Journal of Psychohistory, 9, 199-236.