Introduction: This post is an experiment in synchronization. Since Henry Jenkins, Beth Coleman, and I are all writing about Second Life and because we like each other’s work, even when (or especially when) we disagree, we’ve decided to all post something on Second Life today. Beth’s post will appear at, and Henry’s is at

Let me start with some background. Because of the number of themes involved in discussions of Second Life, it’s easy to end up talking at different levels of abstraction, so let me start with two core assertions, things that I take as background to my part of the larger conversation:

  • First, Linden’s Residents figures are methodologically worthless. Any claim about Second Life derived from a count of Residents is not to be taken seriously, and anyone making claims about Second Life based on those figures is to be regarded with skepticism. (Explanation here and here.)
  • Second, there are many interesting things going on in Second Life. As I have said in other forums, and will repeat here, passionate users are a law unto themselves, and rightly so. Nothing I could say about their experience in Second Life, pro or con, would matter to those users. My concerns are demographic.

With those assertions covered, I am asking myself two things: will Second Life become a platform for a significant online population? And, second, what can Second Life tell us about the future of virtual worlds generally?

Concerning popularity, I predict that Second Life will remain a niche application, which is to say an application that will be of considerable interest to a small percentage of the people who try it. Such niches can be profitable (an argument I made in the Meganiche article), but they won’t, by definition, appeal to a broad cross-section of users.

The logic behind this belief is simple: most people who try Second Life don’t like it. Something like five out of six new users abandon it before a month is up. The three month abandonment figure seems to be closer to nine out of ten. (This figure is less firm, as it has only been reported colloquially, with no absolute numbers behind it.)

More importantly, the current active population is still an unknown. (Call this metric something like “How many users in the last 30 days have accounts more than 30 days old?”) We know the highest that figure could be is in the low hundreds of thousands, but no one other than the Lindens (and, presumably, their bigger marketing clients) knows how much lower it is than this theoretical maximum.

The poor adoption rate is a form of aggregate judgment. Anything bruited for wide adoption would have trouble with 85%+ abandonment, whether software or toothpaste. One possible explanation for this considerable user defection might be a technological gap. I do not doubt that improvements to the client and server would decrease the abandonment rate. I do doubt the improvement would be anything other than incremental, given 5 years and tens of millions in effort already.

Note too that abandonment is not a problem that all visually traversable spaces suffer from. Both Doom and Cyworld serve as counter-examples; in those cases, the rendering is cartoonish, yet both platforms achieved huge popularity in a short period. If the non-visual experience is good, the rendering does not need to be, but the converse does not seem to be true, on present evidence.

Two Objections

There have been two broad responses to skepticism occasioned by the Linden population numbers. (Three, if you count ad hominem, but Chris Lott has already covered that.)

The first response is not specific to Second Life. Many people have recalled earlier instances of misguided skepticism about new technologies, but the logical end-case of that thought is that skepticism about technology is never appropriate. (Disconfirmation of this thesis is left as an exercise for the reader.) Given that most new technologies fail, the challenge is to figure out which ones won’t. No one has noted examples of software with 85% abandonment rates, after five years of development, that went on to become widespread. Such examples may exist, but I can’t think of any.

The second objection is a conviction that demographics are irrelevant, and that the interesting goings-on in Second Life are what matters, no matter how few users are engaged in those activities.

I have never doubted (and have explicitly noted above) that there are interesting things happening in Second Life. The mistake, from my point of view, is in mixing two different questions. Whether some people like Second Life a lot is a completely separate issue from whether a lot of people like it. It is possible for the first assertion to be true and the second one false, and this is the only reading I believe is supported by the low absolute numbers and high abandonment rates. Nor is this an unusual case. We have several examples of platforms with fascinating in-world effects (Alphaworld, Black Sun/Blaxxun, The Palace, Dreamscape, LambdaMOO and environments on the SuperMOO List, etc.), all of which also failed to achieve wide use.

It is here that assertions about Second Life have most often been inconsistent. Before the uselessness of Linden’s population numbers was widely understood, the illusion of a large and rapidly growing community was touted as evidence of Second Life’s success. When both the absolute numbers and growth turned out to be more modest, population was downgraded and other metrics have been introduced as predictive of Second Life’s inevitable success.

A hypothesis which is strengthened by evidence of popularity, but not weakened by evidence of unpopularity, isn’t really a hypothesis, it’s a religious assertion. And a core tenet of the faithful seems to be that claims about Second Life are buttressed by the certain and proximate arrival of virtual worlds generally.

If we had but worlds enough and time…

It is worth pausing at this junction. Many people writing about Second Life make little distinction between ‘Second Life as a particular platform’ and ‘Second Life as an exemplar of the coming metaverse’. I would like to buck this trend, by explicitly noting the difference between those two conversations. I am basing my prediction of continued niche status for Second Life on the current evidence that most people who try it don’t like it. My beliefs about virtual worlds, on the other hand, are more conjectural. Everything below should be read with this caveat in mind.

With that said, I don’t believe that “virtual worlds” describes a coherent category, or, put another way, I believe that the group of things lumped together as virtual worlds have such variable implementations and user adoption rates that they are not well described as a single conceptual group.

I alluded to Pointcast in an earlier article; one of the ways the comparison is apt is in the abuse of categorization as a PR tool. Pointcast’s management claimed that email, the Web, and Pointcast all were about delivering content, and that the future looked bright for content delivery platforms. And indeed it did, except for Pointcast.

The successes of email and of the Web were better explained by their particular utilities than by their membership in a broad class of “content delivery.” Pointcast tried to shift attention from those particularities to a generic label in order to create a club in which it would automatically be included.

I believe a similar thing happens whenever Second Life is lumped with Everquest, World of Warcraft, et al., into a category called virtual worlds. If we accept the validity of this category, then multi-player games provide an existence proof of millions-strong virtual worlds, and the only remaining question is simply when we arrive at wider adoption of more general-purpose versions.

If, on the other hand, we don’t start off by lumping Second Life with Warcraft as virtual worlds, a very different question emerges: why do virtual game worlds outperform non-game worlds in their adoption? This pattern is quite stable over time — it well predates Second Live and World of Warcraft, as with first Ultima Online (1997) and then Everquest (1999) each quickly dwarfing the combined populations of Alphaworld and Black Sun (later Blaxxun) despite the significant lead times of those virtual worlds. What is it about games that would make them a better fit for virtual environments than non-games?

Games have at least three advantages other virtual worlds don’t. First, many games, and most social games, involve an entrance into what theorists call the magic circle, an environment whose characteristics include simplified and knowable rules. The magic circle saves the game from having to live up to expectations carried over from the real world.

Second, games are intentionally difficult. If all you knew about golf was that you had to get this ball in that hole, your first thought would be to hop in your cart and drive it over there. But no, you have to knock the ball in, with special sticks. This is just about the stupidest possible way to complete the task, and also the only thing that makes golf interesting. Games create an environment conducive to the acceptance of artificial difficulties.

Finally, and most relevant to visual environments, our ability to ignore information from the visual field when in pursuit of an immediate goal is nothing short of astonishing (viz. the gorilla experiment.) The fact that we could clearly understand spatial layout even in early and poorly rendered 3D environments like Quake has much to do with our willingness to switch from an observational Architectural Digest mode of seeing (Why has this hallway been accessorized with lava?) to a task-oriented Guns and Ammo mode (Ogre! Quad rocket for you!)

In this telling, games are not just special, they are special in a way that relieves designers of the pursuit of maximal realism. There is still a premium on good design and playability, but the magic circle, acceptance of arbitrary difficulties, and goal-directed visual filtering give designers ways to contextualize or bury at least some platform limitations. These are not options available to designers of non-game environments; asking users to accept such worlds as even passable simulacra subjects those environments to withering scrutiny.

Hubba Hubba

We can also reverse this observation. One question we might ask about successful non-game uses of virtual worlds is whether they too are special cases. One obvious example is erotic imagery. The zaftig avatar has been a trope of 3D rendering since designers have been able to scrape together enough polygons to model a torso, but examples start far earlier than virtual worlds. In fact, visual representation of voluptuous womanhood predates the invention of agriculture by the same historical interval as agriculture predates the present. This is a deep pattern.

It is also a pattern that, like games and unlike ordinary life, has a special relation to visual cues (though this effect is somewhat unbalanced by gender.) If someone is shown a virtual hamburger, it can arouse real hunger. However, to satisfy this hunger, he must then walk away from the image and get his hands on an actual hamburger. This is not the case, to put the matter delicately, with erotic imagery; a fetching avatar can arouse desire, but that desire can then be satiated without recourse to the real.

This pair of characteristics — a human (and particularly male) fixation on even poorly rendered erotic images, plus an ability to achieve a kind of gratification in the presence of those images — means that a sexualized rendering can create both attraction and satisfaction in a way that a rendering of, say, a mountain or an office cannot. As with games, visual worlds work in the context of eros not because the images themselves are so convincing, but because they reach a part of the brain that so desperately wants to be convinced.

More generally, I suspect that the cases where 3D immersion works are, and will continue to be, those uses that most invite the mind to fill in or simply do without missing detail, whether because of a triggering of sexual desire, the fight or flight reflex (many games), avarice (gambling), or other areas where we are willing and even eager to make rapid inferences based on a paucity of data. I also assume that these special cases are not simply adding up to a general acceptance of visual immersion, and that finding another avatar beguiling in a virtual bar is not in fact a predictor of being able to read someone’s face or body language in a virtual meeting as if you were with them. That, I believe, is a neurological problem of a different order.

Jaron Lanier is the Charles Babbage of Our Generation

Here we arrive at the furthest shores of speculation. One of the basic promises of virtual reality, at least in its Snow Crash-inflected version, is that we will be able to re-create the full sense of being in someone’s presence in a mediated environment. This desire, present at least since Shamash appeared to Gilgamesh in a dream, can be re-stated in technological terms as a hope that communications will finally become an adequate substitute for travel. We have been promised that this will come to pass with current technology since ATT demoed a video phone at the 1964 World’s Fair.

I believe this version of virtual reality will in fact be achieved, someday. I do not, however, believe that it will involve a screen. Trying to trick the brain by tricking the eyes is a mug’s game. The brain is richly arrayed with tools to detect and unmask visual trickery — if the eyes are misreporting, the brain falls back on other externally focussed senses like touch and smell, or internally focussed ones like balance and proprioception.

Though the conception of virtual reality is clear, the technologies we have today are inadequate to the task. In the same way that the theory of computation arose in the mechanical age, but had to wait first for electrics and then electronics to be fully realized, general purpose virtual reality is an idea waiting on a technology, and specifically on neural interface, which will allow us to trick the brain by tricking the brain. (The neural interface in turn waits on trifling details like an explanation of consciousness.)

In the meantime, the 3D worlds program in the next decade is likely to resemble the AI program in the last century, where early optimism about rapid progress on general frameworks gave way to disconnected research topics (machine vision, natural language processing) and ‘toy worlds’ environments. We will continue to see valuable but specific uses for immersive environments, from flight training and architectural flythroughs to pain relief for burn victims and treatment for acrophobia. These are all indisputably good things, but they are not themselves general, and more importantly don’t suggest rapid progress on generality. As a result, games will continue to dominate the list of well-populated environments for the foreseeable future, rendering ineffectual the category of virtual worlds, and, critically, many of the predictions being attached thereunto.

[We’ve been experiencing continuing problems with our MT-powered commenting system. We’re working on a fix but for now send you to a temporary page where the discussion can continue.]

Leave a Reply

Your email address will not be published. Required fields are marked *