A Tiny Use Case approach to a preliminary overview of formal and stylistic transformations of character designs between the 1990s and 2000s

This is a guest post written by Oscar García Aranda, who is currently a PhD Candidate and pre-doctoral researcher in the Department of Translation, Interpretation and East Asian Studies of the Autonomous University of Barcelona (UAB), and who undertook a short research stay at the JVMG project with us in Stuttgart to work on further developing his dissertation research.

Oscar is also a member of the InterAsia research group and the Centre d’Estudis i Recerca sobre l’Àsia Oriental (CERAO), and works as a lecturer in the UAB East Asian Studies degree program thanks to a FISDUR scholarship from the Generalitat de Catalunya / AGAUR.

His research deals with manga and anime as contemporary Japanese artistic languages, and his doctoral work is focused on drawing an overview of the visual, stylistic and aesthetic culture of shōjo and shōnen manga of the late nineties and the first decade of the 21st century through the examination of character designs. His dissertation also aims to contribute to the elaboration of a formal and visual historiography of manga and anime.

During his research stay with the JVMG project he developed a Tiny Use Case focusing on his research of the transformation of character designs between the 1990s and 2000s. The following is an abridged version of his final report highlighting his research approach and some of his most interesting findings.

A Tiny Use Case approach to a preliminary overview of formal and stylistic transformations of character designs between the 1990s and 2000s (by Oscar García Aranda)

Building on the findings of previous Tiny Use Cases as well as my own research direction, this TUC focused on four initial research questions, that revolve around the formal and historical dimensions of contemporary Japanese visual culture in terms of manga and anime character designs:

  • Trying to define what exactly these manga-anime formal and stylistic conventions consist of.
  • Which are the changes that these attributes experience over time.
  • Which are the “main” ones in each period.
  • Finally, try to identify some reasons or factors behind these formal and stylistic changes.

Based on these initial questions, I proceeded to elaborate a hypothesis by creating a sheet in which different character formal traits (hairstyle, eyes, lines…) were allocated, offering a range of definitions and descriptions on how they would look like chronologically. More specifically, I focused on a time interval that goes from 1986 to 2010, separating five different periods of five years each. The selected time interval is related with the research questions, starting from the assumption that between the late 1990s and the early 2000s we can identify a remarkable formal and aesthetic change in terms of character designs. For this reason, and in order to elaborate a rough preliminary analysis, I manually observed and formally analyzed the main cast of characters from the most popular anime series of each five-year period, and made a rough comparative analysis in terms of formal and stylistic attributes. It is important to note that the selection of the series in this first hypothesis generating phase of my tiny use case was based on my own knowledge and assumptions regarding manga and anime.

Through this exercise, different preliminary results and considerations appeared:

  • There seems to be a sharp distinction between 1990s and 2000s art styles. This shift seems to be quite abrupt, which led me to think about the possible causes behind these changes and even to propose the idea that maybe they are related to certain technical innovations in the case of anime series.
  • Some art styles (such as the highly expressive and deformed art style of series like Slayers or Saber Marionette) are surprisingly confined to a few years or a particular series, a fact that could also suggest the influence of a specific character designer or artist.
  • On the other hand, I also detected the prevalence of certain formal and aesthetic solutions depending on the genre or demography of the series independently of their release period. For example, sparkling eyes are mainly found in shōjo series, while there is a tendency towards more facial traits that signal race and ethnicity in adult series with a more realistic aesthetic.
  • A structural difference in the representation of certain archetypes and moe attributes was also detected, as well as gradual changes of the beauty standards of both male and female characters. The differences in how female characters are sexualized between and within periods could also be related with a change in consumption patterns among the otaku subculture and the wider audience.

Following this initial hypothesis generating phase, a more rigorous approach was taken to actually analyze the changing character designs in anime over these five periods. First, a random sample was taken from anime for each period, making sure that the chosen works were adaptations of manga (as this would be important for developing my work towards a comparison between manga and anime character designs in the future) and not part of the most popular works from that period. The reason for this was twofold. First, my initial hypothesis generation was conducted on series that I had deemed to be the most representative for each five-year period, which meant that they were often the most popular anime from those periods, and I wanted to examine other works from each period as a way to potentially compare my hypotheses with the results of the analysis (although it is important to note that my research design did not follow a strict hypothesis testing setup, nor was it meant to do that, rather this potential comparison is something that could help me generate further ideas going forward). Second, by separating out the top ten most popular anime for each period and making sure that they could not be part of my sample, I could in the future conduct the same analysis I did with my sample on these top series, and then compare the two sets of findings. Maybe there are differences in the stylistic features of the most prominent works and anime in general.

In order to establish a list of the ten most popular anime for each period I used the member numbers on the MyAnimeList website as an indicator of popularity. And in order to generate my random sample of anime, I used the advanced search interface of the AniDB website, where it is possible to define periods and then ask for a random title from a given period. In this way my sample was created title by title, which allowed for checking whether a given title is in the top ten most popular titles or is already a part of the sample (in both cases it was discarded and a new random title retrieved) and whether it is actually an adaption of a manga or not (in the latter case the title was again discarded and replaced by a new random unit). Another upside of this incremental approach to generating my random sample was the ability to gradually increase the sample size.

I also decided on analyzing five characters per selected anime, with main characters being the focus of the research. The five characters were randomly selected from the main cast, and in case there were less than five main characters in a given title (which meant they would all end up in the sample automatically), further characters were randomly selected from the whole cast to reach the required five characters per anime.

The final piece of my research design was the master code table. I developed a set of detailed characteristics that would be analyzed for each character – in part based on my preliminary analysis of the anime that I had deemed to be the most representative of each period as described earlier – and tried to come up with a list of potential values for each of these characteristics. This master code table was further developed during the coding of the characters in my sample.

Following the above described incremental approach my sample started with five series per five year period, analyzing five characters from each series, which gave me a total of 25 series and 125 characters analyzed. After the initial analysis of the first 125 characters, we detected a fruitful progress in the research despite some flaws and problems, and in order to comprehend more data and offer more detailed and veracious results, we decided to expand the sample two more times: first to 10 series per period and 250 characters, and finally to 15 series per period, concluding the sample with a total of 75 series and 375 characters being analyzed.

The following are some of the interesting first results that emerged from my analysis of the characters in the sample. The longitudinal change in the age and gender distribution of my sample followed the patterns identified by Martin Roth and Zoltan Kacsuk in their work on applying the virtual census approach – inspired by Williams et al. (2009) – to Japanese visual media. Some of the stylistic shifts over time I found in my sample – for example in the changes in the predominance of certain face and nose forms – are not actually a result of overall stylistic changes in character design, but rather reflect the changing demographic make-up of the character populations by period.

Instead of these kinds of shifts in stylistic hallmarks that are less about actual changes in character design and have more to do with the shifting demographic make-up of the population of anime characters over time, in the following I will focus on changes that seem to be more closely related with a chronological progression in character designs. One important example of these kinds of shifts to emerge from my analysis was the longitudinal change in the types of textures and color palettes employed for the characters.

Changes in the distribution of textures for hair over time.

Taking a look at the above figure, we can identify a steady increase of light shade textures for hair from the beginning of the 1990s until becoming stabilized in the middle of the 2000s, while the lack of textures decreases sharply during the 90s until recovering slightly during the 2000s.

Parallel to this phenomenon regarding hair textures, in the figure below we can see how the color palette transformations experience quite similar fluctuations during the same time interval, with industrial palettes reaching their zenith in the middle of the 1990s until becoming marginal again in the second half of the 2000s, when plain palettes (an updated version of the pale-dead tones common in 1980s productions) start to proliferate among anime productions.

Changes in the distribution of color palettes over time.

The pattern seems clear: the rise of industrial color palettes precedes the hegemony of light shades as the main hair texture trend and produces the drop of non-texture hairs in the 1990s, while this same pattern recovers when plain palettes start to become dominant for anime productions at the start of the 21th century. Based on this data, we can propose that rather than being framed by gender or age constrictions, textures and color palettes in character design changed over time thanks to different technical innovations and transformations in the making of the anime works, a factor that is also clearly present in most popular series of each period.

However, this preliminary conclusion may be limited to anime only if we take into account the media specificity of both manga and anime, the extensive use of color in anime and how both media use different drawing techniques and materials, leading to different textures and palettes for the same characters in the two media. In any case, the changes in textures and colors are visible and the relation of these fluctuations with technical changes seem clear enough at least in the case of anime series and adaptations. Moreover, we can detect a certain correlation of this phenomena with fluctuations of aesthetics over time that reflects interesting insights.

Distribution of character aesthetics by period.

As we can see in the above figure, expressive aesthetic appeal (understood as a “standard” mangaesque aesthetic in between cartoonish and realist aesthetics usually involving a clever use of angular lines in order to offer the character a certain solidity and volume) tends to decline over time, while the early 2000s sees the emergence of the light synthesized aesthetic, a lighter version of the expressive aesthetic that reduces the use of lineal details offering a flatter visual result, sometimes involving slender shapes and bodies.

Comparing this imbalance with the fluctuations reflected by textures and color palettes, we can also suggest that at least in the case of anime productions the configuration of an expressive aesthetic involves the use of industrial palettes and light shades to a certain extent, while a light synthesized aesthetic would be better reflected through plain palettes and the lack of textures. In fact, this chart also reflects the isolated position of some aesthetics, such as the exaggerated expressive aesthetic in the mid 1990s. Exaggerated expressive aesthetics can be defined as an extreme version of the expressive aesthetic that involves the extensive use of quadrangular and geometrical body shapes with a certain superdeformed appeal that is common for series like Slayers or Saber Marionettes, having then a reflection in some of the characters included in this sample. The fact that this aesthetic is mainly located in the same period interval as the zenith of color palettes might suggest an interesting interrelation between both phenomena, and maybe also offer us a representative overview of the visual formal trends of this particular period.

Finally, I would like to draw attention to one of the interesting points that emerged in relation to the question already explained regarding the employed research design of this tiny use case: are the most popular series for a given period truly representative of the stylistic trends of a given period, or is it rather the case that they skew our perception of what character designs are like due to their immense popularity while being far from representative of their respective periods? The chart below based on period show us one such example regarding hair shape.

Changes in hair shape trends over time.

As we can see in the figure above, smooth and later smooth angular for the last two periods are the two dominant hair shapes, while the rest remain in a lower level of use with their own dynamics (such as the gradual decline of sponge types after the 1980s). Among these other hair shape styles, the low numbers of spiky hair stands out as particularly interesting. Contrary to the popularity of some best seller series that popularized this hair style such as Dragon Ball, it actually remains a minor hair shape if we take a look at the wider context of the manga and anime industry. Interestingly, however, there is a correspondence between the climax of Dragon Ball Z, the zenith of exaggerated expressive aesthetics and the industrial color palettes, as well as with the a certain decline of the smooth hair shape. In relation to these phenomena, we can suggest that the smooth angular type is actually a reflection of the “heritage” that spiky hairs left in the scene of manga and anime character designs, not only by its sudden increase during the late 1990s, but also for its correlative decline after the emergence of plain palettes and light synthesized aesthetics. This finding is therefore a very interesting lead for further investigations in relation to the effect that major hits can play in defining later stylistic trends in character design, while not necessarily being the most representative style-wise for their own respective periods.


Despite the lack of enough data to better support our results, the outcomes of my research are enlightening enough to point out some interesting formal and stylistic fluctuations. The charts filtered by period point out sudden changes in the time intervals between the late 1990s and the 2000s, a phenomena that matches my initial hypothesis regarding the most mainstream works. Some aspects of character design related to texture and coloring are apparently more strictly related with time progression due to the fact that they probably involve major changes in terms of techniques. Finally, some of the outcomes seem to question some of my initial preconceptions and general assumptions, opening the door for a deeper critical research on such aspects.

Most of my findings should be expanded with a greater sample in order to better validate them and to obtain a more exhaustive overview. Different explanations based on manga and anime literature and the dynamics of the industry also need to be further examined and developed. However, in order to really understand the underlying reasons behind formal and stylistic changes, a deeper research must be done focusing on particular formal categories and involving a theoretical framework on manga and anime studies in combination with primary sources regarding different mangaka, creatives and character designers.


This research stay has been funded by the Agència de Gestió d’Ajuts Universitaris i d’Investigació (AGAUR) from the Generalitat de Catalunya through a fund bag for research internships of the FI-SDUR (2022) scholarship.