Please be aware that this article is protected by copyright© law, Title 17 U.S. Code.
(The Psychological Institute of the University of Berlin)
All of the studies presented in this paper deal with the problem of reproduction, but only the second section deals with this problem specifically. The studies presented in the first section were conducted afterwards in order to delineate the so-called retroactive inhibition and associated phenomena regarding superfically similar subject areas, which are the subject matter of the second section. The concept of retroactive inhibition belongs to a specific system of older ideas and methodologies. Thus, in the first studies, lists were memorized and subsequently tested, as was common practice 20 to 30 years ago.
This methodology is no longer fashionable. Everyone has the feeling that learning, remembering, forgetting, reproducing, and recognizing should be studied in specific special situations, and that the results will therefore be of only minimal significance for general psychology. Without a doubt, these methods deserved criticism by Poppelreuter and others. The importance of the studies of Poppelreuter, Kühn, and Lewin is not underestimated. However, the older methodolgy has not fulfilled all of the duties that it should, since the nature of onesidedness and the specificity of particular situations has not been elucidated. This goal has not been fully accomplished.
The criticism that the memorization of series of neutral items is a meaningless activity for subjects is also insufficient. Subjects will participate in such tasks as long as they are convinced that some gain can be made for psychological research, even if they cannot and are not allowed to see it. After all, we do not want to fool ourselves: millions of people remain in the same work situations day after day, even though their tasks are no more meaningful than experimental tasks. One would hardly criticize the classical Gedächtnispsychologie (psychology of memory) for being too far removed from everday experience, just because the subjects were engaged in meaningless tasks. This is a rather general criticism: all psychological experimentation will at times venture into the realm of extreme circumstances and thus deviate from the common course of events. Oftentimes this is the only manner in which scientific decisions can be achieved.
Nevertheless, it is an important distinction, if one planfully arranges extreme and otherwise rarely occurring circumstances in one experiment, or if all experimentation in one domain rests entirely upon the operation of extreme circumstances. The latter occurred in the classical Gedächtnispsychologie, which is the subject matter of the first section. In the first section, it will be clear that the classical method of memorizing and retaining learned material constantly interrupts the influence of intensive countereffects, and that above all the subject matter must make the proceedings a nuisance for the participants.
Naturally, what is more important than such methodological criticisms are the questions regarding the nature of the countereffects, their relationship to the concept of retroactive inhibition, and the connection with the problem of the Gestalt.
When one examines and attempts to memorize a series of 8 pairs, it is one's subjective impression that a single pair is easier to memorize than are 4 pairs of nonsense syllables. It remains to be determined if the subjective opinion will be supported by objective proof, and if this is the case, if the poorer memory of nonsense syllables is due to the frequency of their presentation or due to the quality of nonsense syllables in and of themselves.
To address the second question, the subjects are presented not only with one list, but with 5 serially presented lists (under comparable circumstances and with sufficient time lapses between the presentation of lists). There are five different types of material; thus, 5 lists are presented to complete the cycle. Every type of material occurs in isolated form (as a single pair) in 4 of the 5 lists and occurs in massed form (as 4 pairs) in 1 of the 5 lists, with the content of the pairs varying. According to the statistical method of hits-and-misses, the probability of a hit (correct recall) in the isolated condition should equal the probability of a hit in the massed condition. This statistical comparison is accomplished by using a group of subjects and calculating the total number of hits for the group.
laf -- rig
# -- +
dok -- pir
89 -- 46
red square -- green square
zül -- dap
S -- B
tög -- fem
The following numbers are the result of such a statistical comparison of 5 lists using a group with sample size of 4 subjects.
The instructions prepare the subject for the later task of recall. The pairs were arranged successively as is indicated in the example and were presented for 2 seconds. After a list was presented 3 times, the experimenter engaged in conversation with the subject for 6 minutes. Then the subject would be asked to recall the presented information by writing it down. 25 minutes later the subject would be asked to memorize the next list. Thus, the total amount of time required to complete the 5 lists was only 2 1/2 hours. Those responses which are recognized as hits (correct recall) are not only those which are accurate reproductions of the presented material, but also those which involve only slight deviations from the original. (For instance, in the case of the syllable "küp", the responses "köp" and "küb" would also be considered hits).
Even though the total numbers are rather small -- the highest number of hits achieved by the group of 4 subjects in either the massed or isolated conditions is 16 -- the result is clear. The number of hits is higher in the isolated constellation than in the corresponding massed constellation, regardless of type of material.
If one adds together all of the M-hits and all of the I-hits, it is clear that the M-pairs result in little over 40% accuracy of recall, while the I-pairs result in almost 80% accuracy. It is therefore concluded that information presented in columns in which the types of material are massed is more difficult to memorize or remember (or both at the same time) than if it is presented in isolated form among pairs of different types of material.
A test of this result is necessary for several reasons. It is possible that there are differences in the type of material such that the I-cases are favored or the M-cases are disadvantaged. Thus, control studies, in which the types of materials are varied, are necessary. Effects of fatigue and practice effects also need to be considered and eliminated by varying the order of presentation, such that the lists containing isolated cases are presented serially. Because the lists were presented so closely in time, it is possible that one list influenced the items of another. This effect is eliminated by conducting the experiment with different subjects across different days. For these reasons, the first group of associated series was followed by the presentation of 4 lists of the same sort. Every list was repeated with a new group of subjects, such that 5 lists were presented to 22 subjects. Table 2 shows the results of these 5 lists.
The total number of possible cases for the M- and I-constellations of each condition is the product of the number of subjects times 20 (4 types of material, 5 lists), i.e. 440. 189 hits occurred in M, 328 in I.
Thus, the total result of every later condition corresponds with the first condition in terms of direction. In the 25 individual cases, the number of M-hits exceeded the number of I-hits only once (with the categories and types of material being the same).
This was the result for letters in condition IV, and it is plausible that letters, of which there are only a limited number, are not the ideal type of material for such lists. In fact, greater amounts of interference could have been expected. However, even colors, whether they were memorized as visual material or as nuances of colors, elicited more I-hits than M-hits.
Afterwards, the subjects struggled not only because they had to memorize and associate nonsense syllables, but also because there were many nonsense syllables which were massed, and this arrangement deviated greatly fron the preceding lists.
The first condition has already been discussed. The second is differentiated from the first only in that the I-pairs are replaced by M-pairs and vice versa. In the third condition, only one list is presented per day. Because this leads to a significant increase in the number of hits for both the I- and the M-cases, it may be assumed that the presentation of one list after another leads to the interference of one series on another, which is greatly reduced or eliminated in the third condition. As will be apparent later, this does not argue against the procedure used in condition I, II, and IV, but rather serves as a confirmation of the principle. Condition IV differs from I, II, and III in that every list is presented in its entirety, and that the subjects view them according to the beat of a metronome. In condition V, the time of presentation for each pair is reduced to 1 1/2 second, and the time lapse between the presentation and test of recall is increased to 25 minutes, during which a pseudo-experiment of perceptual psychology is conducted. The time between the single lists is increased in condition V as it was in condition III. The number of presentations varied across conditions from 2 to 3. The material of the M- and I-paris was the same in conditions I and II, while the material was reversed in conditions II, IV, and V. All of these modifications of experimental procedures did not, as the table indicates, affect the difference between M- and I-constellations.
The validity of these findings and their dependence on the quality of the items (the ease with which the different types of materials are memorized) is particularly evident when the hits of M- and S-constellations are compared across different types of materials within a list rather than within one kind of material across different lists. It is not at all obvious, even though the same result is achieved using this methodology; rather, it assigns greater meaning to the factors under examination as compared to the differences between types of material. For example, one of the lists contains 4 letters and 4 isolated pairs of other materials.
It remains to be determined if the number of hits in the I-constellations are greater than the number of hits in the M-constellations within the same list. Table 3 presents the results.
The difference between the 2 constellations is so much more pronounced than the difference between types of materials used, such that the I-cases fail to outweigh the M-cases in only 2 out of the 25 lists. This occurred in lists I and in lists II, the lists in which colors and letters were stacked, respectively; in list I, the number of M-hits was greater than the number of I-hits, and in list II, the number of M-hits and I-hits were equal. If one considers the small number in general and the small number of possible hits in each individual series, then this result speaks for the validity of the effect of stacking.
It should not be assumed that the type of material has no effect on the number of hits; this effect is a secondary one and is more apparent in the individual results (see Table 3). For example, the number of hits for the color pairs is significantly greater than for the parts of figures (see Table 2, last row). In addition, the transition from I- to M-constellations appeared to affect some materials more than others; for example, numbers were affected more than syllables.
In condition VI, the principle of conditions I-V was applied in stricter form. There are obvious concerns with regard to the use of letters and colors. Thus, these pairs were excluded in the 3 lists of condition VI. At the same time, the differences between I- and M-conditions were increased for the remaining materials (syllables, figures, and numbers). Every list contains one type of material in 6 pairs and the other 2 types in one pair a piece. The others are arranged in different positions among the pairs of the same type of material across lists.
The presentation time for each pair is 1 1/2 seconds, and the number of repetitions equals 3, with several days in between the presentations. Every individual constellation occurred in 2 different individual list in the whole series, the recall of which was tested after 6 minutes for one and after 40 minutes for the other. However, the results for these 2 time intervals are not presented separately in the table below. 12 subjects took part in condition VI.
The number of hits are found in Table 4. Because the possible number of hits for M-pairs was three times as great as for I-pairs,  the actual number of hits for M-pairs was divided by 3 to allow for comparison to the I-combination.
If the interpretation of lists I-V holds, and if the effect of massing of identical materials can be augmented, then the differences between M- and I-combinations should be even more pronounced in this condition (VI). This did, in fact, occur, as one can see by comparing Tables 2 and 4, and for each one of the 3 types of material that were used in I-V and IV. The percentage of hits for all 3 together equals 25% for M and 87% for I. The I-cases are thus three times more advantaged (more easily remembered) than the M-cases.
It is predicated that the predominance of one combination over the other also remains not only when comparing the M- and I-cases of identical materials in different lists), but also when comparing the M- and I-cases of different materials within the same lists. The calculation follows (again reducing M-values for the purpose of statistical comparison):
No matter the type of material, the number of hits is quite small when the items are presented in massed rather than in isolated form.
The purpose of conducting condition VI was not only to examine the effect of massing, but also to demonstrate another influence. One must ask oneself if massing interferes with the process of memorization or if it interferes with the stored memory. When the learning process has been completed, and only a trace of the list remains, do the same interfering effects of massing continue to operate as they did in the learning process? To address this question, subjects were instructed to memorize 2 of the lists from condition VI, one of which was tested after 6 minutes, the other after 40 minutes. There was no difference, however, between these 2 conditions; the larger time interval did not serve to reduce the number of hits at all. This question must also be addressed using different time intervals.  It is sufficient for our purposes that similar conclusions are drawn with regard to this subsequent interference (a massing effect in the trace field). In Tables 4 and 5, the results for both time intervals are summed together.
The degree of massing achieved in condition VI resembles the homogeneous series of syllables typical of earlier experiments. Under these conditions, a very powerful interference due to massing of identical materials was observed. This is important for our evaluation of earlier experiments and leads back to the earlier comment. One could assume that the task of subjects in such experiments consisted of artificially creating an association between neutral components like syllables. Considering the aforementioned result, there was certainly no lack of "association" between several syllables (or figures, numbers, etc.).
The subjects in the earlier experiments certainly had to expend a great deal of effort, particularly because of the opposing far-reaching associations resulting from the homogeneity of lists, which served to interfere with the processes of memorizing and remembering. The effort of the subjects probably consisted of the characterization, individualization, etc. of the individual items and groups of items of such altogether monotonous proceedings. The descriptions of what actually occurs during memorizing fit very well with this idea. On the whole, it should be apparent that the use of nonsense constructions is not the only demanding situation in such experiments. When such homogeneous and monotonous lists are selected for statistical purposes (to obtain equivalent associations and potential hits), then the experiments become particularly difficult.
The increased difficulty of learning longer lists was probably explained by the "limitations of consciousness." However, the process is not captured by such an explanation. The opposing forces that the subject must overcome are only so powerful, because the items are of the same type of material.
Experimental results such as those described above could have been anticipated since 1905, when Ranschburg , after conducting many experiments, formulated the hypothesis that the memorizing and retaining of lists was impaired due to the similarity of items. Interestingly enough, these experiments were largely ignored. 
Müller and Schumann  attempted to eliminate the similarities between individual syllables within lists, because they anticipated facilitating or debilitating effects depending on the situation. These authors, like Ranschburg, assumed that the essential similarity was similarity due to equivalence of parts. Thus, Ranschburg used "homogeneous" and "heterogeneous" lists of several syllables, such that the consonants of syllables were repeated in a few lines and not repeated in the other case. This yields objective well-defined levels of similarity, while at the same time distorting the experiments; this occurs because such uniformity of items in the homogeneous case may be recognized by the subjects and may, in fact, improve the learning of the homogeneous series.
In fact, Ranschburg himself eventually discovered this. A further complication of the Ranschburg experimens in that the memorized material is tested twice either with only one or two presentations. Thus, it is difficult to trace the course of the experiments; it is possible that the results of later tests of memory are not independent of the specific experimental history, and that the conditions are not equivalent. A test of recall assesses the traces of lists, which may not be equivalent for homogeneous and heterogeneous series.
In any case, Ranschburg showed that homogeneous series are more difficult to memorize and more importantly are more difficult to remember than heterogeneous series.  The difference is more pronounced when the time lapse between memorizing and recall is greater as opposed to when it is only of a minute's duration. This finding would be more significant if it could be replicated using a more sound methodology.
Theoretically, it would be important if the joining of 2 already well-memorized heterogeneous partial lists leads to interference if the items of the partial lists are similar, thus, constituting homogeneity according to Ranschburg's procedure. The attempt to treat them as one whole lists led to a decrease in accuracy of reproduction. Particularly noteworthy is the fact that the partial lists appeared to recover from this interference over time. However, this finding requires further investigation.
In the next section, it will become clear that heterogeneous lists are not merely a number of I-cases (our terminology). Our previous I-cases can also not be considered pure in this respect.
The studies presented thus far require reproduction by the subjects. A very different and less frequently used approach requires the subject to identify particular iterms as familiar or unfamiliar. It is questionable whether these 2 approaches yield the same results.
We undertook experiments involving recognition using a similar procedure as the one used in the experiments involving reproduction. The series, however, consisted not of 8 pairs, but of 15 items without any indication of pairing, with 3 being I-cases and 12 being M-cases. Only 4 types of material were used, thus only 4 lists were presented and tested, in which one after another each type of material occurred in massed form.
The 3 I-cases were dispersed among the M-cases, such that an I-case did not occur in either the first or last two places in the series.
The subjects in these experiments were 15 persons, who were divided into 3 smaller groups, such that the order of presentation of massed combinations of individual items could be varied. The entire experiment took place in one morning. The time between presentation and testing was 6 minutes and devoted to conversation. 25 minutes lapsed between testing of one list and learning of the next. Each list was presented as a whole with all items presented individually one under another. The subjects went through the items under the guidance of the experimenter (who followed the beat of a metronome). Each item was allotted 1 1/2 seconds. Every list was presented only once. At the beginning of the test, the subjects were informed that they would be shown several syllables, numbers, etc., some of which would be from the previous series and some of which would be unfamiliar. Each time the subject had to decide whether the object was familiar or unfamiliar or that they were unsure. The familiar were counted as +-cases, while the unsure were counted as unfamiliar cases.
Since the time of G.E. Müller, tests of recognition have involved the presentation of only a part of the actually presented materials plus several distractor cases which resemble those materials. The reasons for this procedure are surely important. However, we had to deviate from this procedure due to the potential for interference. Just as there is an effect of stacking and similarity, it is also possible that such effects are present in tests of recognition, and that tests with newly added items are not neutral tasks.
It had to be decided then how to conduct the experiment without any unfamiliar objects in the recognition test when the subjects had been informed that unfamiliar items would be present in the recognition situation. It was apparent after only one presentation that subjects failed to recognize the same series when it consisted of a larger number of items presented in a different array, and that only a few subjects were able to recognize that the list consisted of identical items after this had been suggested as a possibility.
Table 6 presents the results of these experiments in numbers of hits. DI refers to the number of isolated hits of different materials in the same series, and SI refers to the number of isolated hits of the same materials in different series.
For the purpose of statistical comparison, the M-values are divided by 4. There is a total of 180 possible cases.
It is clear from the table that the overall percentage of hits does not even reach 60%. Thus, even though no unfamiliar items were included in the test of recognition, this did not affect the results.
50% are identified correctly for the M-cases, while 68% are identified correctly for the I-cases. This is no extreme difference, yet the difference is in the same direction as with the test of recall, requiring reproduction. This difference is even more salient when comparing the percentages across the 3 groups: 1) 43 vs. 68%, 2> 54 vs. 68%, 3) 55 vs. 67%. The difference is again in the same direction, although the total number of possible hits is reduced to 60 per group. The most careful control is the percentage in the last column; every one of the 4 M-values is smaller than the associated I-value, regardless of whether the DI or SI value is used. (In this comparison, the total number of hits decreased to 45).
By comparing this table to Tables 2, 3, 4, and 5, it is clear than tests of recognition yield smaller differences that do tests of reproduction. The ratio of M- to I-items was 12:3, while it was 4:4 or 6:2 in the series requiring reproduction. Nevertheless, the reproduction experiments yielded a difference of 43 and 75% and 25 and 87%, respectively.
Thus, it is necessary that the ratio of M- to I-cases be high so that strong effects (of massing or isolation) are obtained in the tests of recogntion, as is the case with tests of reproduction (see p. 312).
Due to a theoretical connection, which will become evident later, the tests of recognition place I-cases in a less preferential position than M-cases than in tests of reproduction; however, there is no doubt that I-cases still have priority over M-cases. A second experiment conducted with 2 groups of subjects in a mass testing situation yielded the same results as the first. The lists were set up a little differently than before. There were only 2 which were made up of syllables and numbers. In one, 3 syllables were arranged in the fourth through sixth positions, while the remaining 15 positions contained numbers. In the other, the situation was reversed. The number of subjects in each group (and, thus, for each series) was 14. The results are presented in Table 7.
In the I-series of syllables and numbers, there are 42 total possible hits, while in the M-series, there are 210; thus, the M-cases were divided by 5 for the purpose of statistical comparison.
In this experiment, one cannot really speak of I-cases in that the 3 items of one type of material follow one another and are presented among 15 different items. Instead, this is a relative isolation, but, for simplicity's sake, the I- and M-classifications are retained. The same difference between M- and I-cases is achieved in this study of recognition as in the aforementioned studies. If one compares the values in the first and fourth columns and the values in the third and second conlumns, then one can see the difference between the different materials in the same lists (same subjects).
In each list, three I-cases occurred at the seventh and ninth places, and the three items of the M-materials followed immediately thereafter; in the other lists, they appeared as I-cases. If one compares the I-cases in one list with the identical S-cases in the other lists, the following percentages are obtained: 81 and 74% for syllables, 81 and 45% for numbers.
The recognition experiment was repeated with an extreme difference between massing and isolation. This involved 2 lists of 20 items. One list included 19 syllables and one number, the other 19 numbers and one syllable. The isolated item was in the middle of the list in both cases.
The items in each list were presented serially on little cards, each of which was presented for 1.5 seconds. On each day of the experiment, only one list was presented and tested; the test was conducted 10 minutes after a single presentation. The 10 minutes constituted a free period; the subjects were merely instructed not to think of the memorized list. The order of presentation of lists varied across subjects.
In the test, the items were presented in a different order; unfamiliar items were not added. There were 12 subjects.
Table 8 presents the results with the M-values reduced for the purpose of statistical comparison.
The difference between the number of hits in the M- and I-combinations is not only in the expected direction, but significantly greater than in the previous experiments requiring recognition. 
This is consistent with the extreme differences between massing and isolation in the lists, which far exceeds that of the reproduction experiments (where the greatest difference was 6:2 for M:I).  The results of the reproduction experiments using similar extremely different lists remain to be determined.
The terms "massing" and "isolation" as they were used in the first section are not adequately (precisely) defined. When one attempts to clarify these, a formulation of the experimental findings emerges, allowing the crucial functional factors to become salient. We begin with the term "isolation." According to the manner in which we used the term before, which is in accordance with Ranschburg's usage, one could think that a pair (or individual item) in a list is in the preferred "isolated" position when that particular type of material only occurs within that pair (or item) in the list. Nothing further needs to be said about the other pairs (items) in the list except that they are different from the isolated ones. Such a line of thinking fails to clarify an important point, as is illustrated below. For example, a list consists of 10 pairs with 1 syllable pair and 9 number pairs. The syllable pair is in the third position. The difference between the 2 types of material is considerable. The syllable pair is thus defined as an I-case, and the numbers are defined as M-cases. If one were to substitute one of the numbers with a figure pair, then there would be 2 I-cases and 8 M-cases in the series, since syllables and figures are both distinct from numbers. In this manner, one could substitute each number pair with a different type of material one after another until the last one was replaced. In the end, then the difference between all other pairs among themselves and the syllable pair is equivalent to the initial difference between syllables and numbers. For example:
A B C D E F G H I K
These letters should illustrate the relative difference between types of materials, such that more distant items are equally different from one another as are 2 adjacent items.
Using our terminology, such a list is composed entirely of I-cases.
We will assume that C represents the syllable pair, and that K represents the represents the remaining number pairs. The series with which we started would thus be transformed in the following manner:
K1 K2 C K3 K4 K5 K6 K7 K8 K9
K represents the different number pairs and C (the syllable pair) represents the only I-case.
Now, according to our assumptions, the C in the first series (A B C D . . . K) is as different from the other items as is the C in the second series where it is presented among K's. According to this line of reasoning, the syllable pair C in the first series should have the same preferred position (be at the same advantage) as it is in the second series where it is among K-pairs (numbers). In both series, C would function as an I-pair. However, it is not obvious that this is the case. Rather, it is probable that C is functionally very different in the 2 series, depending on whether C is obviously different from the other items, or if all the items are of the same type of material. If an experiment revealed different results for the C's in the 2 rows, then one could argue that other factors besides the difference between one item and the other items are important, namely the arrangment of the remaining items. The functional nature of C would, thus, be dependent on the form of the whole series.
An analogous case from the field of perception clarifies this point. In the series presented in Figure 1, no item is more salient than any other. Thus, the series is characterized by figures which are definitely distinguished from one another, but no 2 items are more similar or more different that any other 2. Thus, the series is not adequately described by stating that there are differences between the individual items. Rather, it is important that there is equivalence between differences between items.
Thus, a distinction is made between first order (differences) and second order (equivalence) relationships. One could also say that in such a series, there are no flaws (breaks) in the "continuum of differences."  Our example illustrates that no subgroups are formed under such circumstances. (It is assumed that there are no other differences, for example, in the amount of space between items).
On the other hand, a break in the continuum of differences (or similarities) results in the break-up of the series and the formation of subgroups. Figure 2 is an example thereof. The same difference between the rounded diagonal form and the neighboring triangle occurs in Figure 1, namely between the third and fourth items. In Figure 1, these 2 are presented among other items which are all equally dissimilar.
In Figure 2, however, all of the other items are somewhat similar, and only when one gets to the critical item is there a leap in degree of relatedness. The result is that the more similar items form a subgroup, and the critical item is singled out. This is another instance of a second order relationship affecting the perception of a row.
Another almost banal example illustrates this even more obviously. A number of gray circles of equal size are arranged at equal distances from one another. They are all out of the same gray paper with the exception of the third circle which is of a significantly darker shade of gray. The series is, thus, broken up by this darker circle, and the darker circle is singled out at first glance.
A second series is made up of the same darker circle and one of its neighbors from the first series plus several other circles of other gray tones which are arranged to the left and right of these 2 circles. The colors were selected and arrangec such that the whole series would look like a representation of the black-white spectrum, or a part of it, in which each circle is equally differentiated from the previous circle. Examination of this series would not reveal any subdivision; in particular, the critical circle would no longer be more salient that the rest. This is the case even though the difference between the critical circle and any other circle in the series is either as large as or greater than the difference between the critical circle and the others in the first series. However, there is still an appearance of a uniform whole, because there is no break in the pattern of similarity, that is, in the second order relationships. The example shows that equivalent changes in the similarity between items operate in the same fashion as equivalent relationships between all items. , 
The application of these principles to the experiments described in the first section suggests that the findings may not be due to the relatedness between items in terms of type of material, but due to the subdivisions (subgroupings) within the series. That is, M-cases may be functionally disadvantaged because they belong to a subgroup, and I-cases may yield higher rates of reproduction, because they remain independent (isolated) in the list.
This question may be addressed by comparing the aforementioned series:
A B C D E F G H I K
K1 K2 C K3 K4 K5 K6 K7 K8 K9
When pairs are singled out like this, this surely enhances their reproducability. C, the syllable pair, is equally differentiated from every other member in the 2 lists. However, the 2 lists differ in terms of their pattern of similarity (or dissimilarity) in that the second series is characterized by a break between C and its neighbors K, in that C is singled out, whereas the C in the first series is integrated into the whole series. From now on, we must distinguish between 3 typical cases:
If an arrangement of equally different items of the same type reduces the reproducability of items, then the first case must produce the lowest number of correct reproductions.  In the second case, on the other hand, they will be particularly high.
The third case will be more difficult to memorize than the isolated case even though the items are objectively different from one another, because the continuous great difference between items creates the impression of a uniform whole. That is, case 3 may be seen as a type of "homogeneous" case.  Naturally, the same results are not expected for case 1 and case 3. The strength of a field is dependent on the size of the difference between several items in that field. The heterogeneous series is uniform like the homogeneous one, but it is more loosely associated. Thus, the heterogeneous series will produce lower numbers of accurately reproductions than the true isolation case, but higher numbers than the pure homogeneous case.
Following this discussion of the concept of isolation in a more precise fashion, it is evident that the so-called I-cases mentioned in the first section were not really true to their name. For example, in the first series, 4 items of the same type were arranged next to 4 other pairs which were different from the adjacent ones as well as from each other in terms of type of material. Such a series is a rather confusing yet atypical mixture, in which the first 4 pairs formed a homogeneous field, while the adjacent pairs were not isolated, but rather heterogeneous. At the same time, there is yet another complication which takes away from the functional quality of the homogeneous group. The later experiments (p. 304ff. and p. 312), which involve only 2 or 3 types of material, more closely resemble the typical cases. However, experiments with purely heterogeneous lists have not yet been conducted. The results of these in comparison with the homogeneous and I-cases will reveal whether the pattern of similarity within lists determined the results, i.e. if the formation of fields (in the Gestalt sense) is instrumental.
A first attempt of this nature used the following procedure:
Each subject memorized three lists:
1 number, 9 nonsense syllables
1 syllable, 9 numbers
1 number, 1 syllable, 1 color, 1 letter, 1 meaningful word, 1 small photograph, 1 symbol, 1 button, 1 punctuation mark, 1 name of a chemical compound
List 1 and 2 contain one isolated element and 9 M-elements. In one, the syllables were massed, in the other numbers. The purpose of this duplication is to compare the I- and M-constellations for the same type of material. List 3 is heterogeneous; both the syllable and the number should, however, be seen as critical items, i.e. the results thereof are comparables to those from series 1 and 2. Each list was presented only once.
The 10 items within each list were presented on little cardboard cards of equal size. The individual cards were presented one after another to the beat of a metronome (presentation time = 1.5 seconds). The subjects were instructed to attend to the list and to memorize each individual item. The I-items and the syllable and number which occurred in the heterogeneous series were traded off several times, so that the results would be independent of the particular materials. A space of at least one day separated the learning of the 2 lists.
In lists 1 and 2, the syllable and number were presented at the beginning of the lists, namely at the second and third positions, at which point the subjects could not know anything about the contents of the whole list. Thus, the isolated item was not perceived as unusual and was not particularly salient to the subjects. We wanted to avoid the situation where the critical item would stand out as perceptually unique. Thus, the homogeneous items had to follow the critical items in the lists 1 and 2, so that they could be excluded from the field only after they were no longer visible. For the same reason, the subjects learned the purely heterogeneous list 3 first, in which the critical items played no special role, which could have led to biased perception in 1 and 2. Its position in the list was approximately the same as in series 1 and 2. When the presentation of one list was completed, the subjects were instructed to memorize as much of a meaningful text as possible. After 10 minutes, this task was interrupted, and the test of recall was conducted using the "method of retained items." That is, the subjects had to write down everything that they could remember from the earlier list.
After 30 seconds, the test was completed (and the subjects were asked to recall the meaningful text and write it down as accurately as possible).  15 subjects participated in these experiments. The experimental conditions were the same for all.
The numbers are summed up in Table 9, and the M-cases are divided by 9 so that they are comparable to the I-cases.
The value for the M-cases is very low (22% accurate reproduction), whereas those for the I-cases in lists 1 and 2 are over three times as high (70%); the finding discussed in the first section is replicated here using a different method of assessing recall and a different type of presentation.
Perhaps an even larger difference between the 2 constellations was expected, since the massing was so extreme. It is not appropriate to compare the quantitative results from condition VI in the first section to this one, since the lists, manner of presentation, and assessment of recall are so different (compare, however, bottom p. 324).
The right side of Table 9 provides the number of hits for the critical items in the true I-case (70%). Since the differences between the critical items and the remaining ones in list 3 are at least as large as those between the critical items and the items in the homogeneous field in lists 1 and 2, and since the number of accurately reproduced critical items is lower for the heterogeneous series, there is no doubt that the progression of differences within the whole series is decisive.
If this is the criterion for the meaning of the formation of fields in the Gestalt sense, then it may be concluded that our results rest on these formations. As soon as the break in similarity is removed, that is, the number is removed from the syllable series, and the syllable is removed from the number series, then this global connection is obtained and the accuracy of recall is cut in half.
The same result is obtained when assessing the reproduction of the critical items positioned in the same locations (in isolated form in 1 and 2 and as part of the heterogeneous series in 3). During the assessment of recall, the number and kind of numbers and syllables that the subjects wrote down in their first 5 seconds were recorded. The results showed that the subjects wrote down 18 of the 21 I-cases in lists 1 and 2 in the first 5 seconds, but that they wrote down only 6 of the 12 numbers and syllables presented in the heterogeneous list 3. The difference is obvious.
By the way, the number and the syllable did not form the exception in list 3; although it contains a meaningful word, photograph, button, etc., i.e. items which could be considered highly "memorable," the average accuracy of recall from these other items in list 3 is 6.5 (43%), which is practically identical to the number 6 for syllables and numbers. One can see how the individual types of material are less important than the structure of the list.
Finally, the results for the syllables and numbers in the heterogeneous list 3 are comparable to the same types of materials in the homogenesous field in lists 1 and 2. Earlier we stated the prediction that the rate of accuracy of recall would be lower in the heterogeneous series than in the case of true isolation but higher than for the true homogeneous list. This prediction was confirmed. The heterogeneous list yielded twice as many hits as the homogeneous lists 1 and 2. One could argue, following the earlier discussion, that this finding is due to the lack of relatedness of the items in the heterogeneous list, and that dissimilarity of items is only effective as it fails to lead to field formations.
On the whole, results were consistent with predictions: the smallest number of correct reproductions occurred in homogeneous fields, i.e. lists with the same type of content.
In heterogeneous lists between which there is the same degree of dissimilarity, the items are in a more favored position.
The highest number of reproduced items occur for isolated items within a homogeneous field. 
The experimental principle guiding list 3 eliminates the concern about the comparison of M- and I-cases, as is the case for lists 1 and 2. One might argue that numerous syllables and numbers united in one list could result in substitutions, contaminations, and the like, which is impossible in the case of I-cases. Such subsitutions and contaminations do indeed occur in our procedure. However, the results of list 3 (in comparison to the I-cases of 1 and 2) show that our thesis is in no way undermined by the role of the formation of fields with reference to substitutions and contaminations. That is, no one substitues or contaminates button, number, photograph, syllable, and symbol. Yet still the accuracy of recall in the heterogeneous series is significantly less than in the I-cases in 1 and 2.
The syllable and the number, the critical items, may only be neutral items in the heterogeneous list, because this list was presented first. If lists 1 or 2 (or both), which contained only syllables and numbers, had preceded 3, then the subjects would probably have noticed them more readily than other items. Earlier (on p. 319), we mentioned that the critical items were placed in the second or third position in the series in 1 and 2, so that at the moment of perception there would be no clear field, and the critical item would not be viewed as isolated. Repetition of the presentation was avoided. Thus, it is unlikely that the field effects, which determine the dissimilar rates of reproduction of items in different constellations, operate only at the initial perception of the lists. One could assume that similar effects operate in the "trace field" as well, since field formation occurs only after the perception of the item, whose reproducibility is later determined by the field formation (compare p. 306 and 308).
Since the rules according to which the reproducibility of items is a function of the structure of the list are consistent with the laws governing whether individual parts in the field of vision remain independent or integrated into a perceptable whole, a difficult problem arises. Either similar forces are at work in the "trace field" as are in perception, or our results are merely the direct results of Gestalt effects in perception. As long as this is not settled experimentally, the following argument should be heeded: the heterogeneous series 3 is composed of several very different successively presented items. The idea that these (e.g. syllable and number) are rather difficult to perceive, because of the uniform progression of the series and more difficult than the same typed of items in the I-case, is a rather questionable one. We would favor the notion that the detrimental effect of this uniform progression of the series is first evident in the "trace field," producting a decrease in the accuracy of reproduction.
Although the differences in the 3 constellations are quite pronounced, the possible cases for the critical items in isolation and in the heterogeneous series are rather low; thus, this first experiment should not be viewed as decisive.
The lists in this experiment were not presented so that pairs of items were closely presented. The lists in the next experiment were presented in pairs, each one 3 times, and were tested 10 minutes later. Three lists analogous to the ones in the last experiment were used and contained the following.
1 number pair, 7 syllable pairs
1 syllable pair, 7 number pairs
1 syllable pair, 1 number pair, 6 pairs of closely related items which were as dissimilar in material as those in list 3 in the last experiment.
Only one list was memorized and tested per day. There was no planned acticity during the time between memorization and test of recall; the subjects were merely instructed not to think of the memorized lists.
List 3 was learned first by all subjects; afterwards, some of the subjects learned lists 1 and then 2, while the remainder learned 2 and then 1. The individual syllables and numbers which made up the critical items were frequently interchanged. There were 16 subjects (the teachers in a school).
Table 10 contains the results:
The left side of the table confirms the results of the first section. This certainly is one of the most robust findings in this area. In the M-constellation, the syllables and numbers only achieved about 1/5 of the number of accurate reproductions of the I-constellation. This ratio may have been even greater if the number of presentations had not been so high, and the number of hits for the isolation case were not already at the maximum. Insofar as one can compare the different experiments, this result exceeds that of the recognition experiments in Section 1 (Table 8), where the percentages for M- and I-cases were 40 and 96%, respectively.
The most important comparison, namely that of I-cases and the critical items in the heterogeneous list 3, does not yield the predicted differences due to the technical error discussed above. The number of hits, at least for syllables, is somewhat smaller in the heterogeneous series, but the number of hits for numbers is equivalent in the 2 constellations. Obviously the 3 presentations in this experiment worked so well that the constellation in second place was able to outrank the one in first place, achieving a maximum of hits.
List 3 again shows that syllable and number pairs in the heterogeneous list behave in the same fashion as the remaining pairs (91% as compared to 90%). Thus, in this kind of list, the syllable and the number function just like other items.
The number of presentations was reduced in the next experiment, so tht the number of hits for the I-constellation and the critical items in the purely heterogeneous series could be compared statistically.
The test of the critical items in the M-constellation was not conducted, as the comparison between massed and isolated cases has revealed the same results in so many experiments. Thus, we were able to limit the experiment to the presentation and test of only 2 lists. Of these one was constructed like the heterogeneous list in the last experiment (with 2 pairs added), while the other contained 1 syllable and 1 number pair plus 10 pairs of small black and white drawings of familiar objects. Thus, one could argue that the syllable and number pair were not totally isolated in the strictest sense of the term. However, they must have approached the special position of the I-pairs and remained more segregated than the corresponding pairs in the purely heterogeneous list.
The lists were presented only once. The subjects were allowed to converse for 10 minutes after the presentation; then they were tested.
The experiment was conducted with 2 school classes (42 boys, 42 girls). There were 2 groups, for whom the critical pairs in both lists were exchanged. The presentation and test of the heterogeneous list preceded the experiment with the other list by 1 week (compare top p. 322).
Table 11 shows the results.
The number of hits are higher in the I-case, where the critical pairs appear once in the series, than for the heterogeneous series, where they also occur once. This holds for both the syllable and the number. Thus, the results obtained in previous experiments with the test of recalled items also hold for the test using the method of hits and misses. The observed effect is dependent on the field formations in the lists; the effect is not only determined by the similarities between the critical and remaining pairs of the lists.
It is possible that a comparison of pure I-cases with the same type of items in a heterogeneous constellation would have resulted in a larger difference. To make this comparison, 3 different experiments at different points in time would have been necessary. This was not possible using the same school classes.
The result cannot be traced to the difference in relatedness of the critical items and the other items. The heterogeneous list was constructed in the same manner as in the previous 2 experiments. A pair of pictures appeared next to the syllables and numbers, in the same manner as they appeared in the homogeneous part of the other series (compare p. 319). When there was a significant difference between the critical and other pairs, it occurred in the heterogeneous list. However, due to the consistent (uniform) progression of differences this still led to the "quasihomogeneous" effect, i.e. resulted in lower numbers of reproduced items.
One can also not assume that the list with the many meaningful pictures was easier to memorize and remember than the heterogeneous list, and that the subjects thus had more energy left over for the memorization of the syllables and numbers. In fact, exactly the opposite is the case. Since the meaningful pictures occur in true M-constellations, they are much more difficult to memorize, to remember,and to reproduce correctly than the numbers and syllables in the heterogeneous list. The former only resulted in 13% hits, while the latter resulted in 65%.  The small number of hits for the meaningful pictures in the homogeneous domain again points to the fact that the type of material is much less important than the arrangement of the list. One would naturally assume that such pictures would be memorable, yet even the nonsense syllables which appeared in isolated form yielded 5 times as many hits (69%). On the other hand, when the same picture pair occurred in a heterogeneous list, then the percentage of hits was 60% (instead of 13%), which approximates the average value for all items in this series. It is assumed that such a pair of meaningful pictures would have resulted only in hits in isolation, such that the result would have been 100% for the I-case, 60% for the heterogeneous case, and 13% for the homogeneous case. That is the pattern that would be predicted theoretically.
The effect of the perception of the list is also not the only effect operating in this experiment, because the first list which is presented and tested is the heterogeneous list in which the syllables and numbers play no special role. The list in which the syllable and number do play a special role (due to their position at the beginning) follows, and this one is presented only once. Thus, if syllable and number are regarded as special, this is because thy are retained in the trace field.
It is also impossible to trace the results to substitutions and contaminations, rather than the effect of field formation, since the effect is present even in the heterogeneous list in which substitutions and contaminations do not occur (compare top p. 322).
If it is the case that the influence of the structure of a list on memorization and remembering is due to the respective field formation, rather than the similarities (or differences) between individual items, then all of these field formation have the same effect, regardless of the progression of similarity in the list or other factors. In fact, it is obvious that isolation can be accomplished by other means than through a break in the progression of similarity, for instance through separation in time and space (compare p. 328). The placement of individual items in a homogeneous list is not uniform in terms of time and space. The old finding that the beginning and ending items in a homogeneous list are easier to remember than the "middle items" is understandable given that they are regarded as somewhat "more isolated" than the other items.
It remains unclear how items (or pairs) are adversely affected by their perception in uniform fields. From the psychology of perception, we know that figures become parts of a larger field (whole), then changes occur within those figures. However, we are accustomed to perceiving these changes (referred to as illusions) as moderate as long as the field formation does not affect the discreteness of the parts. The items in the lists in our experiments remain discrete while they are presented, i.e. while they are perceived. Thus, the assumption that the significant impairment in our experiment is comparable to the effects of field transformation in perception only holds if the assumption also holds, that such transformations go farther in the trace field (or in the formation of traces) than in the perceptual fields. When we solve this problem, the question of how many accurate reproductions are possible under certain conditions on the basis of the number of given traces is less important than the question of how the traces are altered, such that accurate reproductions are not achieved. 
Also, it is not entirely implausible to conclude from the described experiments that the perception of a figure within a whole (field) always results in impairments (interferences) like the ones described here.
The detrimental field formations which occur in these experiments can be characterized as "monotonous." Thus, a figure which has been inserted and is highly specific can become part of a whole situation. It is unlikely that such an insertion will result in the kinds of impairments discussed above. 
It can also be said that, in the case of recognition, it is not the specific similarities of items in a list, but the formation of fields that determines the results in the different constellations. This was tested, and results were obtained in the predicted direction. However, for technical reasons, the results of this experiment were not considered methodologically sound. A discussion of these difficulties would accomplish little.  It is necessary to develop a more precise methodology for the problem of recognition.
At the end of the preceding section, the reader was reminded that field formation is dependent not only on relatedness, but also on the presentation in time and space. Thus, when all other circumstances are equal, the ease with which components of a field are integrated in a domain (field) and the degree of association of components are dependent on the dimensions of time and space. We apply this rule to our area of research. There are 2 positions on this matter. First of all, the effects may be increased through the quick sucession of items. Secondly, the effects observed in immediate succession would not disappear when the time intervals between items were longer, but merely be observed in a weaker form. Several conclusions follow:
When 2 groups of several items of the same type of material are separated by a lapse in time, this is expected to have the same type of detrimental effect (but to a lesser degree) as the effect of the massing of items presented in immediate succession.
If the detrimental effect operates in both directions in time, i.e. from earlier to later items and vice versa, then one would predict that the memorizability of a number of later items of a particular type would be adversely affected by the fact that a larger number of items of the same type were presented in the not too distant past.
If the detrimental effect of massing is weaker in tests of recognition than in reproduction (compare top p.310f. and 324), then it is predicted that this decrease will be even more noteworthy when the effect is carried across a certain time interval, such that the detrimental effect loses intensity. It is, in fact, possible that hardly any detrimental effect would be recognizable in this case of recognition. This conclusion would hold for the case in which the material of the same type preceded as well as followed the other items.
The first conclusion would involve a retroactive inhibition, which would be detectable in the test of reproduction, such that (contrary to the initial view of G.E. Müller ), the reduction in the inhibitory effect would occur between the presentation of the inhibiting and inhibited items. Such a finding would demonstrate that the formation of functionally related fields and corresponding detrimental effects would operate across a lapse in time. Experiments which attempt to modify the concept "retroactive inhibition" in the described direction have been conducted in America;  these have not always revealed the same results, but a tendency which argues for the necessity of such a reformulation of the concept.
In terms of the second conclusion, we can also draw on American empirical findings.
According to these, it is at least somewhat likely that a "proactive inhibition" of earlier on later items is operating, and that this effect increases with the increasing relatedness of material. 
The third conclusion is consistent with R. Heines' finding,  according to which no "retroactive inhibition" was observed in the test of recognition. This finding is now easily comprehensible in this connection (which follows as a consequence of the results discussed in section I), whereas it was not easily comprehensible earlier. In this respect, experiments are necessary which show that no increase in the relatedness of fields ("primary list" and "list") yields the same definitive "retroactive inhibition" as is found in the test of reproduction. Nothing is yet known about "proactive inhibition" in tests of recognition.
If retroactive inhibition of similar proceedings operates in a simlar fashion as the detrimental effects of massing discussed in the first 2 sections, then the older experimental methods can easily be improved. If one desires a strong inhibition, then one must prevent any other type of detrimental effect of similar intensity from occurring before the inhibitory proceedings are presented. If one list and one comparison list are associated, as they are in such experiments, then a further damage of one of the 2 (through retroactive inhibition) will not be easily detectable. Thus, it is suggested that the material to be affected by retroactive inhibition be presented in a long homogeneous list, as is typical. Only a few items of the particular type should be given such that the learned product is in a favored position and can be affected more strongly by subsequent interfering effects. The failure to recognize this is probably the reason why the inhibitory effects were so weak in earlier experiments. We proceeded in the following manner.
A larger number of subjects memorized the primary list, which consisted of 2 syllables, 2 figures, and 2 number pairs.
Afterwards one of the groups of subjects were presented with 4 secondary lists, which consisted of 6 syllables and 3 numbers each; the remaining subjects were asked to memorize 4 other secondary lists, which consisted of 6 figures and 3 numbers each. Thus, a massing of syllables follows the primary list for the first group. If our reasoning is correct, then the number of hits for syllables in the primary list will decrease in comparison to the number of hits for syllables in another group in which the secondary list contains no syllables. On the other hand, the other groups would yield lower numbers of hits for figures in the primary list, because many figures appear in the secondary lists which are absent in the first group. One should expect similar results for both groups of subjects in terms of numbers, since numbers represent a neutral material in comparison to syllables and figures, for which the same conditions are established in the secondary lists of both groups.
The subjects were students in a girls' school. Each class made up a group of subjects (15 and 13 subjects).
The pairs of the primary list were presented one after another at 2-second intervals, and the whole list was presented 4 times. (a short pre-experiment was conducted with the presenation of meaningful words followed immediately by a test to familiarize the subjects with the procedure of presentation and testing.)
The 4 secondary lists were shown in a similar manner as the primary list, except that they did not contain pairs and were only presented once. After the presentation, the subjects were instructed to write down all of the items they could remember in a given amount of time. The total amount of time between the competion of the memorization of the primary list and the beginning of the test was 8 minutes for both groups.
The following results were obtained in the primary list:
The letters U and G in the table indicate whether the constellatins are disadvantaged (ungünstig) or advantaged (günstig), that is, which of the material would suffer from greater or lesser decreases in accuracy of reproduction according to the type of secondary lists. The columns represent the same tested material, whereas the rows involve different material, but for the same group of subjects.
Both means of comparison lead to the same conclusion: each time the number of hits is smaller for the disadvantaged constellation. The most important difference is that between groups for the same type of material. This difference is quite large for syllables and is also clear for figures. On average, the advantaged constellations yielded 80%, while the disadvantaged ones yielded 48% hits.
The numbers inthe primary list were intended to be neutral,such that in both types of secondary lists, the influence of syllables and figures was equally minimal, and the influence of subsequent numbers was strong. The number of hits are quite low for both groups, but are not comparable, since group I received 22% and group II received 44%. It is not unlikely that such a large difference is attributable to individual differences in the 2 groups of subjects.  It is more likely that numbers are more similar to syllables than to figures, and that the massing of syllables in the secondary lists (of group I) had a greater effect on the numbers in the primary list than the massing of figures (in group II). This assumption will be confirmed in the next series of experiments.
An intensive effect of massing of similar items was also observed in the method of recalled items used in the experiments discussed in the second section. We will prove the last result using this methodology.
The arrangement of this new experiment corresponds to the last one in most characteristics; however, a third comparison condition was added, thus, the time between the presentation of the primary list and the test was used in 3 (instead of 2) different ways. The primary list did not contain 9 pairs, but 9 individual items (the first items in each pair of the earlier argument), and the subjects were required to reproduce the items by writing down all the items they could recall without the assistance of "repoducible moments."
The subjects (48 university students in a lecture class) were divided into 3 groups (16, 15, and 17 people). All 3 groupd viewed the same primary list, which consisted of 2 syllables, 2 figures, and 5 numbers, 3 times.
Afterwards, group I memorized 4 secondary lists of 6 syllables and 3 numbers, while group II memorized 4 secondary series of 6 figures and 3 numbers; group III did not memorize anything, but spent this time solving difficult cognitive problems.  (Each secondary list [group I and II] was presented twice and tested immediately after.)
Each of the 3 groups had the same amount of time (2 minutes) for writing down the recalled items (10 minutes after the last presentation of the primary list).
The symbol GG is added to the symbols U and G described above. GG indicates a particularly advantaged position in terms of our earlier conclusions.
Next we will compare the results from group I and II. This time the values for the G-condition are much higher than for the U-condition whether one examines the columns (same group, different material) or the rows (different group, same material).
An average of 52% correct reproductions were obtained in the G-case vs. 8% in the U-case. There is no longer any doubt that the so-called "retroactive inhibition" is determined primarily by the degree of relatedness of materials in the primary and secondary series. One may conclude that the "retroactive inhibition" is a special case of the massing effect discussed in sections I and II.
If the numbers represented a type of material that is neutral compared to figures and syllables, then the values for accurate reproductions should have been the same for groups I and II. Both of these values are very low, but they are not equivalent. In group I, in which the secondary list contain syllables in addition to numbers, the percentage of accurate reproductions was 11%; on the other hand, in group II, in which figures replaced the syllables, the percentage was 24%. This is exactly the same kind of asymmetry observed in the preceding experiment. Thus, one may conclude that the syllables are more closely related to the numbers than are the figures to the same numbers, such that the syllables (in the secondary list of group I) have a stronger retroactive inhibitory effect on the numbers than do the figures (in the secondary list of group II).
the fact that such differences in relatedness are apparent in the differences in values of correctly reproduced items speaks for the suggested interpretation of retroactive inhibition.
It is quite noteworthy that the values of correctly reproduced items are much higher for group III (the GG-condition) than not only the U- but also the G-conditions of groups I and II. In the time lapse between presentations and testing, the subjects in group III are not concerned with syllables, figures, numbers, not even with lists that resemble the primary lists, but with the solving of meaningful problems, a demanding activity. This means that a maximum of dissimilarity between the primary list and the subsequent task, along with the formation of a separate field in the interim, fail to lead to the formation of a whole field. Therefore, the retroactive inhibition had a weaker effect than in the previous G-cases. On the other hand, one can conclude that differences in terms of material such that figures and numbers had an adverse effect on syllables and numbers and syllables had an adverse effect on figures. thus, there is a parallel between the scale of varying degrees of similarity and the varying degrees of the effect of retroactive inhibition. 
The value for accurately reproduced numbers is quite high for the GG-condition, namely 66%. The fact that this value does not approach the values for the syllables and figures is not attributed to the fact that these are particularly difficult numbers to remember, but rather that the primary list contains more numbers than syllables or figures, and that the numbers are, therefore, subject to a stronger massing effect in the primary list. 
G.E. Müller's  assumption that retroactive inhibition is not bound by the type of material but operates due to cognitive effort (perhaps fatigue) is supported through a case study.
In this case, the objective contents of the subsequent task were unrelated to the memorized material, yet a retroactive inhibition was still observable. In the comparable experiment, the time interval was a free period. Clearly 2 cases are being compared in which both involve limited association (relatedness) between memorized material and the contents of the following period, so that the meaning of the degrees of association for the observed finding is not easily recognizable, particularly for such an experiment. In addition, it is possible that, in the comparison of the GG-condition and another one in which there is a free period, the latter would result in the higher number of hits. This is also what is predicted from our theoretical perspective.
In the first section (see p. 301f.), the experiments involved a set of 4 pairs of the same type of material and a set of 4 pairs of different materials (M- and I-cases). In the first 5 experimental conditions, condition III had a much higher number of hits for both the massed and the isolated cases (see Table 2, p. 302, summary in percentages on the right). For this condition, the time separating the individual experiments (i.e. the time between the test of one list and the presentation of the next) was increased from 25 minutes to at least 24 hours.
This was also the case in condition V, but here the number of hits was somewhat reduced, because the time of presentation for each pair was significantly shorter than for the other conditions.
Retroactive inhibition of later individual lists on earlier ones (conditions I, II, and IV) is not a plausible explanation, since the test follows the memorization of the respective individual lists. The larger the time span separating individual lists in condition III seemed to be advantageous, because a longer period of "list-free" time preceded the memorization of the individual lists.
It becomes clear that the increase in number of hits in condition III is not a coincidence when one computes the average values for conditions I, II, and IV (see the above comment on condition V) and compares them to those for the individual types of material in condition III. This comparison shows that each of the individual M- and I-values in condition III are higher than the average of I, II, and IV.
It is highly unlikely that such a relationship is attributable to other circumstances, like the simultaneous exposition of lists in condition III and the like, of which none resulted in differences between I, II, and Iv. The effect of fatigue still remains plausible, as this was, in fact, operating in the later individual lists in conditions I, II, and IV.
One can also point to such a "proactive inhibition" in these conditions in another manner. Because the individual lists of I, II, and IV follow one another closely in time, these I-cases could be damaged through proactive inhibition, which occurs when the preceding series contained items of the same material in an M-constellation. The following figure illustrates the schema according to which the 5 columns of a condition are arranged:
In the square of symbols, the M-symbols (capital letters) from a diagonal, on either side of which there are the same number of I-symbols, namely 10 that follow the massed position and 10 that precede it. Such a square can be provided for each condition and looks different for the different conditions depending on the manner in which the individual types of material pass through the massed position. If one considers the conditions I, II, and IV together, then one square (like the one above) would show the isolated cases for syllables after the massed list and the isolated cases for colors before the massed list. Table 14 shows the number of hits in the I-constellation before and after the massing.
According to this, the I-cases before the M-constellation yielded many more hits than the I-cases after the massed position. The comparison of condition III with conditions I, II, and IV could only show that, regardless of the type of material, the crowded position of the individual lists reduces the number of hits.
The result of this comparison seems to prove that the proximity (in terms of time) of items of the same type of material leads to interfering effects, which vary according to the degree of relatedness of materials.
To be exact, this evidence is not powerful enough, since fatigue can also be operating here as is proactive inhibition. If one examines the schema of the experiment on p. 336 in which the columns are the combination of the individual lists presented on one morning, one can see that lists 1 and 2 contain 7 of the 10 I-cases which appear before the accompanying massed positions, while in lists 4 and 5, 7 of the 10 I-cases follow the after the accompanying massed positions. On the whole, the I-cases of the second type occur later in the experiment and, thus, are more likely to be subject to the effects of fatigue. It cannot be assumed, however, that the difference between the results in the 2 constellations can be explained in this manner; it is too large for that. An experiment similar to the one used in the investigation of retroactive inhibition is necessary.
If our ideas are correct, then the comparison of conditions III and V, in which there is a large time lapse between the presentation of the individual lists, will not yield such a large difference, or at least a much smaller difference than the comparison of conditions I, II, and IV: 
According to this, it seems plausible tht "proactive inhibition" may operate to a lesser degree over a period of days. If this is the case, then a comparison of Tables 14 and 15 shows that the interference effect is much weaker over longer periods of time. The number of hits for I-cases increased dramatically after the massing. On the whole, this experiment confirmed the result of an American study, which raised the suspicion that proactive inhibition exists.  The massing effect within lists corresponds to proactive inhibition from list to list, which manifests itself as the decrease in number of hits.
The individual lists in the first experiment in which the effects of massing and isolation were studied with tests of recognition (compare p. 329), we concluded that proactive inhibition (with relatedness of material varying by degrees) should be minimal or not occur at all the with the same type of test. This question can be addressed in much the same manner as the case of tests of reproduction, since more groups of subjects also went through the massed constellations for the individual types of material in different orders (compare top p. 336). Table 16 shows the +-cases (with I-constellations) before and after the accompanying M-constellation.
One can no longer speak of a predominance of +-cases before massing. In fact, the small difference may be meaningless.  Proactive inhibition was not evident in this test of recognition.
It remains to be determined whether retroactive inhibition occurs in tests of recognition. The experiments designed to address this question were modeled after the corresponding experiments with tests of reproduction.
A larger group of subjects was shown the primary list once, which consisted of 2 syllables, 2 figures, and 5 numbers. Afterwards one group of these subjects was instructed to memorize 4 secondary lists presented one after another; these consisted of 6 syllables and 3 numbers.
The remaining subjects were instructed to memorize 4 secondary lists consisting of 6 figures and 3 numbers. That is, the presentation of the primary series was followed by the massing of syllables but not figures for group I and massing of figures but not syllables for group II. The numbers were assumed to be neutral items.
93 subjects (university students in a lecture class) participated, of which 46 belonged to one group and 47 to the other. The items in both the primary and secondary lists were presented one after another for 1.5 seconds. Following the presentation, subjects were instructed to write down all the items they could remember in a certain period of time. The time interval between the end of the presentation of the primary list and the beginning of the test of recognition was 8 minutes long.
The test of recognition of syllables and figures of the primary list  yielded the following results:
The letters U and G in the table have the same meaning as before. The comparison within rows, which involves the same material, yields a difference favoring the G-condition for both types of material. This difference, however, is so minimal that one cannot argue that retroactive inhibition occurs in a test of recognition.
In the tests of reproduction, it was determined that numbers were not neutral materials as compared to syllables and figures, but that they (numbers) are more easily "inhibited" by syllables than by figures. Since the test of recognition for syllables and figures failed to reveal a significant retroactive inhibition even with related materials (Table 17), then one would naturally not predict that the numbers would be more strongly affected by syllables more than by figures in these circumstances. The test of numbers yielded 54% in group I (followed by syllables) and 49% in group II (followed by figures).  Thus, there is, in fact, no difference between the 2 groups.
Since we repeated similar experiments and never found a predominance of hits in either direction (using sums), we can conclude that the simple connection discussed above (p.329) holds: recognition is less affected by the massing effect than is reproduction in the strict sense (section I). One may assume that a reduction in the massing effect through a separation of the materials in time leads to a faster decline in interference (damage) in test of recognition than in tests of reproduction. If retroactive inhibition merely involves the stacking effect across an interval of time, then it will no longer be evident for recognition, while it will still be prominent for reproduction. This is confirmed in the experiment.
Another question remains, that is, whether the interference (damage) is weaker in the tests of recognition than in the tests of reproduction in the massing experiments (section I). This question is easily answered: recognition is a less specific type of function. Alone the repetition of a relatively similar character among others which deviate dramatically leads to the perception of the relevant object or series as faamiliar. Thus, the trace of an earlier series can be significantly damaged without the recognition of the familiarity of a series being damaged. Let us reflect on the test methodology: at least a trace of the pair which is to be reproduced must be differentiated to such a degree that one can carry this from its basis to the subdivision such that the "right" process is utilized. Otherwise, a mistake is made. With the test of recognition, it is possible that a significantly damaged trace still resembles the original enough that the relevant item is identified. The same trace would be insufficient as the basis for a specific reproduction.
A part of this line of thinking is found already in Woodworth and Pottenberger [Textbook of Experimental Psychology 1920, cited by van Ormer, Psychol. Bull. 30, No. 6 (1933)] as an explanation for Heine's results: the task of recognition is perhaps not sensitive enough to reveal concrete differences between primary and comparison lists, that is, to detect retroactive inhibition even if it is present.
Although this conclusion is quite plausible and is supported by the findings in our experiments that there is a relatioship between the massing effect and retroactive inhibition, this has not been proved through these experiments.
This is due to methodological limitations which argue against the use of typical tests of recognition in general and their application in our specific case. When we present someone with a pair of nonsense syllables and then after a while present them with another one of these syllables, then this one will without a doubt be recognized as "another one of those syllables." It is questionable whether this type of general familiarity in the course of the experiment is significantly different than the kind required in the test of recognition, namely the recognition of a particular individual syllable. If this strong separation (differentiation) is not guaranteed, then one can make the following objection to the tests of recognition discussed in the first section: in addition to the specific familiarity effect, there is also a general one, which is related to the type of material. With regard to the general effect of familiarity, massed material shoud be more advantaged, because of its massing, than the other types of material that have only one item per series. This can result in a large number of hits in the test of material that was massed in the presentation, which is attributable to the general effect of familiarity, while a similar advantage of the so-called I-cases is only expected to occur to a lesser degree. Therefore, the difference in the results for the I- and H-constellations, even when it benefits the I-constellations (as is the case in the test of reproduction), is probably smaller than would be the case with flawless methodology. Thus, it is possible that the massing effect found in the test of recognition was smaller merely due to the weakness of the methodology.
A similar phenomenon is observed with respect to the question of retroactive inhibition in the test of recognition. When the primary list is followed by secondary lists that contain many related itmes, then this will lead to a substantial general effect of familiarity. In the event that retroactive inhibition actually operates in the case of tests of recognition,then this effect would be compenstated at least in part by the opposing effect of familiarity of a general type. When Heine used materials which were not exactly of the same type (as the primary series) in her "inhibitory series," she provided a good reason, even though the decrease in relatedness should render the strongest inhibitory effects ineffective.
The same holds for proactive inhibition. In this case, it is at least theoretically possible that our results of tests of recognition also deviate due to this secondary influence.
One could attempt to eliminate the effects of these methodological flaws through the introduction of new items (similar in material) in the test. It is, however, not certain tht this kind of assistance would achieve this purpose. Yes, one must consider whether the same error would not be made again. Probably the new test items would also have an effect on the results of the test of the objectively familiar items (compare top p. 309). It is easy to see that this influence would have a differnt effect on I-items in a series than M-items.
Thus, even this route does not seem practiable. Only when we find a totally different kind of methodology which is equivalent to tests of recognition but requires much more specificity, can we make a final decision about the two possible explanations. By the way, this insecurity only exists for the case of tests of recognition. The results of the tests of reproductiuon with their specific performance demands remain unaffected by these concerns.
There are intensive forces operating in the monotonous lists of similar material used in classical studies of the psychology of memory; these tend to reduce the established effects of learning. Thus, items which are not presented in such a monotonous massing achieve much higher values of accurately reproduced items than those in massed positions. This detrimental effect is not only based on the conglomeration of similar items, but also on the field formation and the absorption of items into fields, which benefit from the uniform progression of lists. Tests of recognition rather than reproduction lead to the same result, albeit to a lesses degree. Retroactive inhibition and proactive inhibition are principally parallel forms of the same interference effect. The lesser degree to which the field effects appear to operate in the test of recognition is attributed to the absence of retroactive and proactive inhibition in the same type of test.
(Submitted on September 20, 1933.)
(Submitted on September 20, 1933.)