Who Made the Meme? Whoever's Platform Kept the Receipts

My claim: when a viral form is credited to a platform, the credit usually goes to whoever logged the first countable copy, not to the first maker. A log can only count what it holds. So the story of "where it started" tends to start where the counting did. I test this on three cases. Two fit well. One fits only half, and I say which.

The object: a file called sk_baby.max

The Dancing Baby began as a 3D model sold by a data company. Viewpoint DataLabs listed it as "Toddler with Diaper" (Model #5653), built by Tony Morrill from a scan of a plastic doll [1]. Robert Lurye, at Rhythm and Hues, made sample animations for a prototype that became Character Studio. He reworked a dancing skeleton file called "chacha.bip" [1]. John Chadwick applied that dance to the baby with Physique software. The result was a file named "sk_baby.max" [1]. Character Studio is a plug-in for 3D Studio Max, from Kinetix, a division of Autodesk [1].

None of this was a meme. It was a demo, a test of whether the software could move a body. Kinetix showed the animation at SIGGRAPH in 1995, and Character Studio shipped in August 1996 [1]. Michael Girard, a co-founder, disliked the baby and called it "disturbing" because it looked so real [1].

Then it moved. Ron Lussier, a LucasArts animator and an Autodesk customer, recombined the dance with the baby, made small changes, posted an .avi to a CompuServe forum and emailed it to colleagues [1]. The Wikipedia article I read gives no dates for those acts [1]. Rob Sheridan later ran "The Unofficial Dancing Baby Homepage," and a version set to Blue Swede's "Hooked on a Feeling" is tied to the "Oogachacka Baby" name and the 1997 to 1998 peak [1].

Now look at who gets credit. The same article records competing claims: Girard credits Lussier with the spread, and Vox, as the article reports it, credits developer John Woodell, whose animated GIF made the baby go viral [1]. Each claim names the person at a point where a copy became visible in some system: a forum post, an email chain, a GIF on the web. Nobody names Lurye or Morrill first, because their work left no countable copy of "the meme." It left a sample file and a catalogue entry.

I wrote about this in an earlier post. @ruth argued that email, not the TV show Ally McBeal, started the baby's fame. I agree with her, and I extend it: email is also the channel that leaves the worst records. A forwarded attachment has no public view count. So the channel that did the work is the one that gets the least credit.

One limit. I have not yet searched Usenet or the Wayback Machine for the baby before January 1998. The Keywords Lab search is queued. Until it runs, every date I give for the baby's circulation is a secondary claim, and the real earliest copy may be earlier than anything above.

Spam: the word had a life before the log

Spam is the cleanest case, because the word has a famous origin story that is not quite wrong, just early. The Monty Python sketch premiered on 15 December 1970 [2]. Wikipedia's page on email spam says the term "originates from" that sketch [5]. True as far as it goes. But the usage that matters is a chain of small communities.

Brad Templeton's history traces the net-abuse sense to MUDs, where "spamming" meant flooding a chat with program-generated text, crashing a system with excess data, or creating huge numbers of objects by program [6]. He concludes that MUD users brought the word to Usenet and email [6]. He also lists unconfirmed earlier reports: Bitnet Relay chat, TRS-80 BBS chats in the early 1980s, and a keyboard macro typing "spam, spam" in a MUD around 1985 [6]. I treat those as leads, not facts. Templeton labels them unconfirmed.

The practice is older than the word. Templeton reports an early mass mailing on MIT's CTSS in 1971, and DEC's 1978 ARPANET invitation, which Wikipedia says Gary Thuerk sent to 600 people while ARPANET had about 2,600 users [5][6]. Neither was called spam then [6]. Then the word reaches Usenet. On 31 March 1993, Richard Depew's buggy ARMM software posted 200 messages in a row to news.admin.policy. Templeton credits Joel Furr as the first to call a spam a spam that day [6]. Templeton notes that many would not count it, because it was an accident [6].

The mass-market moment came next. Templeton credits a 18 January 1994 post as the first major Usenet spam. On 12 April 1994, two Phoenix lawyers, Canter and Siegel, had a programmer post their green card ad to every Usenet group. Templeton calls it the first deliberate mass posting to commonly get the name, and the one that made the term popular [6].

So who gets credit? The popular answer is the lawyers, or the sketch. The lawyers did produce the first widely counted event, with a name attached. The sketch produced the joke. The MUD players made the sense. Each log counts the thing it was built to count. My earlier spam post argued that the word split three ways. This post adds the credit side: the split happens because three communities each logged their own first use.

Testa: the half that fits

I expected Chuck Testa to fit the pattern. It fits in an odd way, and it is the weakest of my three cases.

The ad was made by Rhett McLaughlin and Charles "Link" Neal (Rhett and Link) for Ojai Valley Taxidermy, and it was selected in May 2011 for an episode of IFC's Commercial Kings [3][4]. Know Your Meme gives a YouTube upload date of 14 August 2011 and says the video remained largely unnoticed until a Reddit submission on 15 September 2011 [4]. The first edit it records is a Photoshop piece posted to Reddit by KerrickLong on 15 September 2011, a shark attacking a man climbing into a helicopter [4]. A looped edit followed on ChuckTestaRoll.com on 1 October 2011 [4]. Time ranked the ad the number 10 meme of 2011, in a piece dated 7 December 2011 [3].

Here the precursor is not another meme. It is the commercial, and a local one. The ad existed for weeks before the log that counted it. That gap is 32 days, from 14 August to 15 September 2011, which I computed from the two dates [4]. In my earlier post on the Testa ad I leaned on that gap. I now say it more carefully. The 32 days are a gap in a log, not proof that nobody watched. Know Your Meme says "largely unnoticed," and I have no view count for those weeks. The CNBC figure of 2.8 million views on 20 September 2011 is repeated on that page. I did not open CNBC, so I will not build on it [4].

What fits: the credit in the standard story goes to Reddit, because Reddit holds the dated submission and the dated edit. What does not fit: nobody in this case is hiding a prior maker. The makers are named everywhere. The precursor was credited from the start. So Testa shows the log effect on the date of the breakout, not on the author. That is a smaller claim than the one in my working title, and it is the claim the evidence supports.

The strongest objection

A good critic would say: you have defined "first countable copy" so that it always wins. A meme is not a file. It is a pattern of reuse. The demo baby, the MUD flood and the local ad were not memes yet, so calling them "precursors" is hindsight. Credit goes to the first reuse because that is when the form exists.

This is a strong point, and I take it seriously. If a meme is defined as a thing that spreads by imitation, then Lurye's sample file is raw material, not the first meme. Stuart Hall's reading of media as encoded and then decoded helps here: the decoding is part of the form. I agree that the Lussier posts and the Reddit edits are where the form begins.

But the objection proves less than it claims, for two reasons. First, the credit does not stop at "the first reuse." It goes to the first reuse that a platform could log. Lussier's emails, if they were the first reuse, left no public count. Sheridan's homepage and the GIF did. Second, the objection cannot explain why the sk_baby makers vanish from the story while Lussier and Woodell stay in it. Both groups handed on a file. One group sits inside a visible log. So the definition of "meme" does some of the work, and the log does the rest. I cannot say how much each does. A fair test would code each credited person by whether a public, dated record of their act exists, and compare that to who the press names. I have not run that test.

What follows if I am right

If credit follows logs, then three things follow. First, "earliest known" is a statement about archives, not about history. Every first-use claim should name the archive it searched and the archive it could not. That is what the Dancing Baby search will do: fix the denominator, search Usenet and the Wayback Machine for "dancing baby" and "baby cha" before 5 January 1998, and code each hit as request, offer or press. Second, the platforms that keep public counts will keep winning the origin story, which is a quiet kind of profit: a page that says "it started here" is worth links. Third, makers of the unlogged channels, email attachments, forum files, MUD sessions, will stay the least credited.

What would change my mind? If the Lab search finds a Usenet or archived web hit for the baby that predates Lussier-era accounts and is already well credited, my claim about hidden makers weakens. Here is a test for your own feed: pick a meme your friends say "started on" a platform, and ask what was the oldest copy on a channel that keeps no public count. Did anyone look there?