Thoughts on Super Mario World 2: Yoshi's Island - 3D Game Design Principles in 2D with Super FX 2

• Developer: Nintendo EAD
• Directors: Takashi Tezuka, Toshihiko Nakago, Shigefumi Hino, Hideki Konno
• Original Release: 1995-08 (JP), 1995-10 (NA, EU)
• Version Notes: This post refers to the original North American SNES release played on a Windows 7 PC via the bsnes emulator (64-bit version 088; new releases use the name "higan"). Xbox360 controller mapped as closely to the SNES controller as possible was used. No video filters used. "Blur" video shader, Direct3D, DirectSound, and RawInput were used.

By the time Super Mario World 2: Yoshi's Island was released for the Super Nintendo Entertainment System (SNES)  to Western markets in October, 1995, both the Sega Saturn and the Sony Playstation (as well as the minor players, the 3DO and Atari Jaguar) had already ushered in the fifth generation of home consoles. With the new generation came the dominance of 3D polygonal games on home consoles, supplanting the sprite-based 2D games that had previously dominated the market. Development of games that use these polygonal 3D spaces is fundamentally different from 2D sprite-based development in many ways. In this post, I will focus on the increased importance of graphical asset deformation as a mode of animation, in contrast to the more frame-based animation of many 2D games of the SNES era. Yoshi's Island, perhaps influenced by the growing 3D development market, implements many of these deformative techniques in its game design, making it one of the most experimental 2D, Nintendo platformers on the SNES. The function of these aspects of its design are likely major aspects of the game's critical success.

Most platformers, and indeed most games, on the SNES achieved animation by moving 2D sprite images relative to each other. Each of these sprites could have multiple frames of animation itself, with the color and location of each pixel in each frame rigidly defined. This approach was prevalent partially due to the SNES hardware, which was practically incapable of rendering complex 3D worlds or even of rotating, scaling, or deforming 2D sprites in-game.  The few games that attempted to do something differently usually had to do so through careful sprite-based trickery. Super Mario Kart (dev. Nintendo EAD, dirs. Tadashi Sugiyama & Hideki Konno, 1992), for example, used a clever combination of the Mode 7 (retrieved 2015-10-14) graphical mode and changing kart sprite size to give the impression of distance in a 3D world. Even the changing sprite size, however, was not done on-the-fly in the game; instead, the developers created the illusion of sprite scaling by swapping sprites of different sizes in and out of the display. (And even this false-3D world required an additional piece of hardware - the DSP-1 chip - on the game's cartridge.) These hardware constraints naturally informed the sorts of game design that could be accomplished on the SNES, and the restrictions imposed by such mostly-static, sprite-based design can be seen in the myriad 2D platformers released for the system. (There were exceptions of course: Clever software rendering on only the native SNES hardware allowed alternate approaches by a handful of games, such as Wolfenstein 3D [devs. id Software & Imagineer Co. Ltd., 1994], which managed a pseudo-3D first person shooter.)

In time, developers sought to expand the SNES's hardware capabilities by including additional processing power on the game cartridges themselves. Perhaps the most famous of these coprocessor chips were the Super FX and its successor, the Super FX 2. Nearly all of the games that used these Super FX coprocessors attempted to use the aging SNES to compete with the next generation of game consoles by creating games in 3D. However, these early attempts at 3D games on the SNES were simplistic; though they managed to capture some simple thrills of flight or racing, they could not explore the promise of 3D for game design in any real depth. Ultimately, these attempts at bringing 3D to the SNES were a dead end from a game design perspective. The one Super FX title that attempted something different with the additional power afforded by the Super FX 2 was Yoshi's Island. It deploys its coprocessor to expand the possibilities of 2D platforming, and as a result it ventured into more creative and experimental game design. The principles that it explores are ultimately closer relatives of later 3D game design than any of its 3D Super FX cousins.

Yoshi's Island is superficially quite similar to its predecessors in the Mario series. Its most obvious departures - the crayon-like art style and the Yoshi-protecting-Mario mechanic - have been amply discussed elsewhere, so I won't dwell on them here. Instead, let's consider how the game leverages the capabilities of the Super FX 2. Throughout the levels, numerous obstacles - including a few polygonal ones - are maneuvered in interesting ways. However, I want to focus on how the design of boss battles has evolved in this game compared to its predecessors, and how those techniques relate to more modern game design perspectives.

A game that takes place in a 3D world has a vastly larger number of possible positions and movements for its assets and camera than one that takes place in a 2D world. This feature precludes any possibility of planning the graphics pixel-by-pixel, as had been done in many sprite-based 2D games. Instead, 3D graphical assets are defined and then deformed, such as into a walk cycle, before the resultant pixels are calculated and rendered to the screen. This approach accommodates a flexibility of design in which a greater variety of player actions can be accommodated. Despite taking place in 2D, many of the boss battles in Yoshi's Island take similar approaches, something that previous Mario games could not do owing to hardware limitations. Where past Mario bosses typically required the dodging of patterned obstacles while trying to jump on the boss's particular weak spot, Yoshi's Island's bosses create variety that encourages critical thinking and dynamic attack strategy while maintaining excitement through the combination of this diversity and stylized graphics. These boss battles control player movement and positioning dynamically by using very large sprites, they allow a diverse set of positions and angles from which the player can attack via deformations and rotations, and they heighten tension during the battles by creating depth via dynamic sprite scaling. Let's look at several examples.

Sluggy the Unshaven deforming
in response to an egg throw.

First, consider the mid-world boss of World 5, Sluggy the Unshaven.  Sluggy is a large, gelatinous creature (reminiscent, perhaps, of the Big Beanie boss previously seen in level 1-8) that fills the tunnel in which the battle takes place from top to bottom, creating a dangerous wall that the player cannot move beyond. In another lovely graphical trick that shapes this scene's structure, Sluggy is transparent and, as the player quickly discovers, houses a vulnerable heart in the center of the gelatinous mass. Initially, the game suggests no mode of attack, but the player cannot remain inactive for long, because the large, impassable, squirming image of Sluggy (not easily possible on native SNES hardware) is forcing the player to move slowly and tensely leftward toward a bottomless pit. Because the blob has no obvious way in which it can be jumped upon like the bosses of past games, the impending doom forces the player to desperately try alternate strategies. Upon throwing an egg at Sluggy, a solution is finally revealed: Sluggy both moves back slightly and, shockingly at the time of the games release, dynamically deforms in response to the egg. The player now realizes that the deformation can provide a way out of this predicatment, because the repeated use of thrown eggs can deform Sluggy into a shape with exposes his vulnerable heart.  The use of large, deformable sprites has created tension which prompts creative player action, which then in turn introduces a completely new game mechanic. Crucially, this attack strategy is subject to the player's creativity. Sluggy can be deformed in any direction from any location; there is no one fixed point of attack as would have been expected in earlier platformers or even many subsequent ones. It is this flexibility, enabled by the Super FX 2, that makes the battle so exciting and memorable.

Raphael the Raven's battle on a
planetesimal, made possible by liberal
use of sprite rotation.

Later in World 5, the player encounters Raphael the Raven in a battle seemingly pulled directly from Nintendo's future, when they sent Mario into space for Super Mario Galaxy (dev. Nintendo EAD Tokyo, dir. Yoshiaki Koizumi, 2007). Like that later Wii title, this battle is able to explore the movement of platforming from a different perspective by translating the jump mechanic to a scenario in which the gravitational field varies spatially. If the Mario series is a mediation on jumping, then this battle, like a musical piece playing with variations on its melodic theme, is an exploration of a variation on the series's movement theme. After the characters dramatically fly up to a moon-like object in an animated sequence achieved with rotation and scaling, the moon serves as the obstacle in a cat-and-mouse game. By clever rotation of the character sprites, their projectiles, and the background, the player must deploy their jumping and ground-pounding skills on the planetesimal, whose gravity always points toward the center of the screen.  Put another way, the developers have asked the player to execute their jumping in a coordinate transformation.  Despite taking place in 2D, the Mario games up to this point have always limited their movement controls and their strategic deployment to what is effectively a 1D space - left and right in Cartesian-like coordinates. The same is true in this battle, but the concept is made refreshing by moving to polar coordinates with the Super FX 2's rotation capabilities - the player still moves primarily in 1D, but now it is the azimuthal angle that the player manipulates.

Baby Bowser looming in the
background, inching forward.

I'll finish off my examples as the game itself finishes: with the final battle with a giant Baby Bowser. Here, Yoshi's Island engages in fairly direct emotional manipulation of the player using techniques analogous to the previous battle with Sluggy the Unshaven. In that earlier scene, the movement of a large sprite leftward created tension that spurred the player to action. Here, the enormous Baby Bowser sprite, perhaps the most dramatically shaded sprite in the game, is scaled up and down to give the illusion of depth into the background. He gradually moves toward the player (both the player-character and the screen itself), all the while destroying the ground upon which the player stands, creating the most tense battle yet. Again analogous to the battle with Sluggy and using the knowledge that the player gained in that battle, the only way to hold Bowser at bay is to throw eggs, this time into the screen, which are scaled and rotated rapidly to give the appropriate illusion of depth.


These battles, and others like them, set the tone of the game as a whole. Yoshi's Island encourages experimentation in strategy on the part of the player by continually subverting the player's expectations of what will be required of them in a Mario game. The game is only able to make this design work because it adopts the sprite-deformation techniques enabled by the Super FX 2 and perhaps informed by the burgeoning 3D game design industry. In this sense, Yoshi's Island is the first truly modern Mario game in game design sense, as it, for the first time, has full control of its graphical assets and can deploy them against its themes. The small segments inform the whole, exploring the mechanics in greater and more experimental depth that had been achieved before.

Despite the critical acclaim that Super Mario World 2: Yoshi's Island and its re-releases received, there has been little critical discussion of how it is mechanically distinct from its predecessors owing to its hardware. Indeed, the word "Yoshi" has never even been mentioned on Critical Distance (retrieved 2015-10-16), despite the game's enduring legacy, both in terms of the sequels it spawned and the impact it likely had on future Nintendo game designs. Instead, it has become lost in the sea of Mario games, and what few consumer-guide-oriented reviews exist for the title largely focus only on surface-level discussions of the presence of an infant Mario character and the crayon-styled graphical assets. Such neglect of any deeper criticism is a shame, because while some aspects of its level design had become run-of-the-mill by the time of its release, other aspects, such as those discussed here, were truly remarkable for their subtle innovations.

A new statement of the scientific consensus on the effectiveness of brain-training games

A while back, I wrote a short review of the scientific results on whether "brain training" games like the Brain Age (devs. Nintendo SDD & Nintendo SPD, various dirs., 2005-2013) series have any cognitive benefits beyond mere entertainment, as their marketing strongly implies. I argued that while it is possible that cognitive decline might be mitigated in certain populations by playing such games, the effect was poorly established and no better (and possibly worse) than other mental exercises, whether they be more traditional activities or any other video game. This question is clearly one that is on the minds of many consumers; search results leading to this post are one of the largest sources of traffic to this blog.

Recently, a group of scientists has issued a statement through the Stanford Center on Longevity that summarizes the current scientific thinking toward such games. The statement, "A Consensus on the Brain Training Industry from the Scientific Community," is available online. It largely agrees with my assessment, arguing that there is no evidence that these games are better than other activities in reducing cognitive decline. It, however, offers new and different citations to the scientific literature that my post did not have access to at the time, and it is written by experts in the field. Please give it a read if you are interested in this topic.

The Brain Age Series: Claims and Effectiveness

I've always been a bit annoyed by games like Nintendo's Brain Age series (also known as the Brain Training series in some regions) for the DS. The marketing for these products is very clear in its implications: playing these games will improve your mental skills and capacity. And yet, to support such a sweeping claim would require a great deal of empirical study that Nintendo has not undertaken. Indeed, Nintendo, despite the implications of their marketing, carefully distances themselves from explicit claims of effectiveness. Nintendo's head of US marketing, when pushed on the matter, affirms that Nintendo is simply "in the entertainment business." This dissonance between marketing, official position, and public perception naturally raises the question of what actual utility Brain Age does or does not offer. I'll try to summarize here what research is available on the matter.

Perhaps it's best to start with the origins of the games. Who is that goofy face that gives you advice on how to play? As one could quickly infer from the name plastered all over the titles, Brain Age was inspired by the work of Tohoku University neuroscience professor Dr Ryuta Kawashima; the game draws heavily from his book, Train Your Brain: 60 Days to a Better Brain. The book is written as a fun self-help guide, and it directly states that users will be able to prevent mental deterioration and will have "more neurons and neural connections." To support these claims, the book simply states that the exercises derive from the "latest research" from Kawashima's "neuroscience lab." Though two simple, qualitative sketches of experiments are given, there are no citations of the actual research. As a result, we'll have to go elsewhere to find out what such exercises really do.

Before we focus on the Brain Age software itself, let's look at more general results. A number of studies have looked at the effects of cognitive stimulation, often in the form of computer-based tasks, on the impact of dementia and/or Alzheimer's in elderly people. For example, Spector et al. (2003) engaged a group of people suffering from dementia with a form of cognitive stimulation therapy consisting of exercises not dissimilar to the Brain Age games. They found improvements in mental state and quality of life due to the exercises that were comparable to those induced by drugs for dementia. Similarly, Belleville et al (2006) looked at a sample of 47 elderly, many with mild cognitive impairment, were treated with "tasks of episodic memory" such as list-recall and association tasks. They found a statistically significant improvement in treated groups over the control group. Other groups have focused directly on the preventative effects of such tasks with respect to Alzheimer's disease. Raúl de la Fuente-Fernández (2006) found that small changes in cognitive activity can result in large reductions in the rate of Alzheimer's in the population studied. It has also been suggested that computer-based tasks can lead to improvements in unrelated tasks such as driving in the elderly. All of these studies, along with many others, suggest that tasks that require cognitive activity, like, presumably, the Brain Age games, can help fend off mental decline as one ages.

Other research has focused on the use of games in people of a variety of ages. Even Kawashima's own research group is involved in an ongoing, but incomplete, study of the use of the game GO as an intervention tool with elementary students. The evidence for positive impact in cognitive ability due to brain training games at other ages is less clear, however. For example, Owen et al. (2010) conducted a six week online study of 11430 participants. They were subjected to a variety of tasks similar to those found in commercially available brain training software.  Though the participants were seen to improve on the tasks that they were performing in their training, the study found no significant effect on other tasks, even if they were cognitively similar to the training tasks. This study casts extreme doubt on the argument that tasks like those provided by Brain Age can improve the mental skills of healthy adults.

But what about Brain Age specifically? Learning and Teaching Scotland conducted a study of 634 Scottish elementary students from 32 schools who used Brain Age daily. They found moderate increases in learning due to the game. Unfortunately, this study does not appear to be peer reviewed. Additionally, the tests used to measure learning gains were only simple arithmetic tests nearly identical to parts of the Brain Age game; a more general test was not used. Further, it is an observational study without a proper control group. As such, these results must be viewed skeptically. Indeed, work by Alain Lieury reports on a study of ten year olds who used Brain Age. He found that students who spent their time on Brain Age performed 17 percent worse than a control group on a battery of tests. Students who used their time on pen-and-paper exercises rather than Brain Age performed 33 percent better than the control group. Lieury extrapolates to adults, suggesting that "if it doesn't work on children, it won't work on adults." Unfortunately, this study, too, does not appear to be available in peer-reviewed form, and all of its details are not available. The sample size is also somewhat small. From what information is available, however, it seems to successfully refute the Scottish study. Kawashima himself has also been involved with studies that investigate Brain Age's effectiveness. One such study compared the use of Brain Age to Tetris in a small sample of 28 healthy elderly persons. Though the study finds a moderate improvement in certain tasks after training with Brain Age, it finds comparable improvements due to Tetris use. In other tasks, however, both groups do not improve and sometimes even do worse.

The net implications of all of these studies seem fairly clear, at least in a qualitative sense. Cognitive activity seems to help stave off mental decline and disease that comes with old age. In healthy populations, whether adult or juvenile, the effects are less clear. Cognitive training may have very modest effects, but no study has conclusively demonstrated a significant one. Brain Age in particular, however, does not seem particularly good at causing any improvements in any population when compared to any other activity. In school children, it may or may not have a positive impact, but, in any case, it does not perform significantly better than any other training activity, including traditional ones. In the elderly, a larger positive benefit is seen, but, again, it is no better than other games, such as Tetris. In short, the literature suggests that there is no compelling reason to play Brain Age over any other game with a goal of improving mental capabilities. Brain Age should be regarded as an entertainment game like any other, unless new evidence comes to light suggesting otherwise. Nintendo's marketing for the game, then, is misleading at best and outright incorrect at worst.