Station to Station.

The real future of Internet radio 

If you spend much time on the Web, you've probably stumbled on some radio stations there - if not one of the unusual operations that broadcast only over the Net, then a traditional AM or FM outlet's site, dedicated to simulcasting the station's signal around the globe. You may also have heard that in just a few years the Internet will be wireless, and we will be able to listen to thousands of such stations on cheap, portable receivers.

There are complications to Net radio, though, that most press accounts ignore, notably the medium's traffic problem. There's a finite number of people who can tune to any audio stream at once; if you top that number - which could be as low as 50 - no one else will be able to hear you. The techies are working on a solution to this, called multicasting. If and when it's in place, Internet broadcasting will be ready to bloom. Until then, though, it will be too bottlenecked to be competitive.

Once multicasting is ready, the Web won't be just a different way to deliver radio shows. Netcasting will converge with at least two other technologies, creating something very different from traditional radio. One of those technologies is the MP3 file, and all the other means of digitally storing music and transferring it over the Net. The second is the sophisticated software that radio stations use to plan their playlists.

It's impossible to predict all the ways this new, hybrid medium will evolve. But by allowing listeners to invent their own personalized radio formats, it's sure to undermine - perhaps fatally - the consultant class that currently decides what gets played on the radio.

Very few radio stations let their disc jockeys choose which songs they'll play; jocks today are paid for their voices, not their musical taste. It wasn't always thus. In the '40s and '50s, the first wave of great R&B and rock 'n' roll DJs didn't just pick their own records; they made radio an improvisatory art form, putting similar songs together, fiddling with the levers and dials in front of them, spinning yarns, making up rhymes, interacting with the music as it played.

This was counterbalanced, in the late '50s, by the rise of Top 40 and the earliest, crudest forms of audience research. Compared with modern pop radio, early Top 40 was spontaneous and far from formulaic. Nonetheless, the transfer of power from jocks to program directors sparked some resistance, and in 1959 several DJs formed the National Disc Jockey Association to advance their interests and protect their privileges.

By the time the group held its first formal convention a year later, however, it was distracted by the first great uproar over payola - that is, record companies bribing jocks to play certain songs.

In that climate, the new DJ association devoted its energies to defending the profession's image, not extending its power. And within the stations, the threat of federal penalties was yet another excuse to revoke the jocks' right to pick their own records.

DJ freedom returned to vogue in the late '60s, as the first "freeform" and "underground" stations appeared on the FM band - staffed, in part, by refugees from the payola scandals. But as the '70s progressed, audience research made a comeback and freeform declined. Today, only a handful of commercial stations have open formats.

The consultants who killed freeform offered a lot of arguments for their approach. The most reasonable was that an audience clearly existed for more narrowly formatted radio, and that it made sense to exploit the new research techniques to find out what those listeners wanted to hear. But if freeform stations risked alienating audiences by tossing in records only a small coterie would enjoy, the formulaic, consultant-driven stations that replaced them overcompensated. Commercial radio today is less concerned with finding music that will draw listeners in than with eliminating music that might drive listeners out. The result is numbing repetition.

One might expect entrepreneurs to take advantage of the situation, by starting new stations willing to take the risks the established outlets won't. In some cases, that might mean a completely freeform approach: Some of us like being surprised by strange sounds we've never heard before.

In other cases, it might simply mean letting DJs examine the research themselves, combine it with their own knowledge of music, and choose the records they play without straying from their station's personality. This was the approach favored by Los Angeles' KMET in the late '70s, after it abandoned full-fledged freeform for a mainstream but still open-ended Real Rock Radio format. It evidently worked, since KMET quickly became the city's top-ranked station.

But under the current regulatory regime, the broadcast spectrum is artificially scarce. There's little room for new stations, and the price of the existing licenses is ridiculously high.

So what does this have to do with the Internet?

Increasingly, program directors are using computer programs to choose the records their stations will play. The most popular such program is called Selector. Once everything in the music library has been entered - not a terribly onerous task, since the typical station has a library of only 500 to 1,000 songs, the vast majority of which are rarely played - the director can give Selector a series of instructions and let the program produce a playlist.

Those parameters might be broad genre restrictions ("no rap"), general patterns ("two upbeat songs, followed by one ballad, then repeat"), or more narrow rules ("no more than three songs with female vocalists per hour"). Selector then chooses which songs will be played, and in what order, for the next 24 hours, seven days, or however far a horizon the program director requests.

Invented in 1979, Selector has produced a series of offspring in the ensuing decade. One is SelectorReach, which combines Selector's data with info from the local Arbitron ratings. ("Filters include age, gender, ethnicity and preference," declares the company Web site, "all of which can be selected individually. For example, you can see black women age 25-34.") Another is Master Control NT, which doesn't just choose which songs to play: It actually plays them, along with all the appropriate ads, promos, and prerecorded DJ bits. Welcome to the completely automated radio station.

Think about this. For decades, each radio station has been trying to figure out the formula that will generate the perfect series of songs for its audience. Now some bright computer programmers have created a tool that will turn its preferences into a playlist. But what happens if its listeners have access to that same program - and to a much larger library of music on the World Wide Web? What if a Web site let listeners select their own parameters and then followed them to the letter, with no commercial interruptions and no DJ schtick?

Some users might choose a familiar generic format, such as those standby oxymorons, "young country" and "classic rock." Some might combine a couple of formats ("play young country and classic rock"). Some might get ridiculously specific ("I like old-school hip hop, mid-tempo ska, country music from before 1970, and Miles Davis' 1959 album Kind of Blue"). It would be a relatively simple matter for the site to track which songs are being played and to pay the appropriate parties a licensing fee, perhaps covered by user subscriptions and perhaps by some other means.

"Some form of that is certainly going to happen," says Tom Zarecki, director of marketing at Radio Computing Services, the company that produces Selector. "It's the same thing as traditional radio programming, but to a more tightened niche than ever before." Already, crude versions of this are beginning to appear. Imagine Radio lets Web surfers create their own stations, with a process that includes checking off as many as 11 musical genres for the station to play. And RealJukebox will shuffle all kinds of digitally stored music: You can load on tracks from your own CDs and you can pull in MP3s from around the Net. The RealJukebox Web site helpfully includes links to hundreds of other sites offering songs for free or for a small fee, many of them far from the musical mainstream.

As the new medium grows, it will continue to evolve. It might, for example, incorporate the technology called collaborative filtering. Such a program would ask you to rate a variety of songs and musicians, and, after accumulating a sufficiently long list and matching it with other people's responses, recommend other songs and musicians you might like. Or, if you prefer, it would just add them to your playlist. (I have experimented with MovieLens, a highly touted Web site that purports to recommend movies on this basis, and in my opinion the program is worthless. Still, the software may yet improve to the point where many consumers would choose to use it. I mention it only as one possible direction the medium could take.)

If Radio Computing Services doesn't offer a music scheduling engine to this market, someone else will: Zarecki estimates that there are about 200 companies in the radio software business, offering everything from scheduling engines to programs that help talk show hosts screen their calls. The only real barrier will be a legal one: The Digital Millennium Copyright Act of 1998 puts a host of restrictions on Net stations to make it harder for listeners to predict when songs will be played, so they won't copy them without paying for them. It's unclear whether this would restrict a customized radio station - nor, for that matter, why such "piracy" is any different from simply taping songs off the radio. If the incumbent radio and record industries get sufficiently scared, though, a legal fight may ensue.

Customized radio won't mean the death of traditional broadcasting. It will mean changes, though, for the research-driven style that currently dominates radio. As Net access becomes cheaper and portable, it will be harder and harder for old-fashioned stations to draw listeners away from online services, especially as they simultaneously face competition from direct-satellite radio, with its hundreds of channels, and from low-power micro-broadcasters (see "Radio Waves," June).

Traditional stations will be able to do one thing, though, that an automated, Web-based system can't. Like the old freeform stations, they can hire skilled knowledgeable hosts who understand how to put disparate songs together in creative sets that no scheduling engine could conceive. A good DJ knows how to introduce listeners to music they haven't heard before, how to put new or eccentric records in a familiar or otherwise enticing context. She might combine this with theater or with interviews. She might invite musicians to perform live in the studio. If she's adventurous, she might make sound collages, mixing different records and other noises together. (This used to be considered ridiculously avant-garde - until someone got the bright idea to put a beat behind it and call it hip hop.)

Then program directors could stop playing super-DJ, and take on the larger visionary role of shaping the station's personality, of figuring out the boundaries of what it will play and finding the right staff to play it.

And if the old stations don't do that, surely there are Web-based stations that will.

Associate Editor Jesse Walker ([e-mail]) is writing a book on micro radio.