Musical Instrument Digital Interface. Baptised in Summer 1983 but born months before following discussions between certain major synth manufacturers seeking a standard for synth connection. (Roland, Sequential Circuits and Yamaha were notable MIDI diplomats. All but a few of the world's synth companies have now joined the embassy.)

The task was to standardise the existing control voltage (pitch) and gate (on/off) information and establish a common language for future developments — programming, memory selection, sequencer control, etc.

Communication would be via five pin MIDI sockets though only two of those pins would be used. Digital information would be sent down the line in serial form — imagine a row of pennies touching each other, tap the first with the edge of a ruler and the last jumps off the line. At the far end the digital bits are converted to flashes of light on an LED like morse code. A phototransistor spots the flashes and converts them back into digital information, that way there's no direct electrical contact between the two synths and hum loops are avoided. The sending speed of the "morse code" is known as the Baud rate, and is set by a crystal clock. It has to be the same in both synths if the messages are to be understood.

MIDI gets off to a good start but it soon becomes obvious that there are strange gaps. Some machines are virtually incapable of understanding each other except on the basic pitch and trigger levels because their technologies are so different. And the synthesiser industry is still suspicious. Would a universal connection system encourage the punter to buy one of their keyboards instead of buying one of ours???

At first it seemed MIDI would unlock the door to boundless new soundscapes. But we'd barely had time to untangle our DIN leads and the whole business had turned into a maze.

It was all becoming distinctly unfriendly out there in the system-exclusive world: unless you owned equipment all made by one manufacturer you were likely to encounter problems. My first attempt was linking a Roland JX3P to a Yamaha DX7. Both keyboards would use MIDI channel one, just plug in two leads and... what could be easier? Well, the JX3P responded and the sound of the DX's vibrant percussive tones mingling with Roland's fat strings was encouraging. But the DX declined to take control, which was less than satisfactory as it is in many ways the ideal "mother" keyboard.

Along with a chosen few from One Two I went along to Syco to investigate MIDI, The Myth. There we found stacks of systems: Sequential Circuits; the new Roland MIDI controller keyboard with modules; a DX7, DX9 and Juno 106 with the Yamaha RX15 drum machine and Roland MSQ700 sequencer. They also had a well-informed operator called Kendall to lead us through the maze. He demystified my failed experiment in a few seconds.

The DX, you see, can't transmit and receive on the same MIDI channel. Therefore you need another piece of equipment (an MX1 which will cost you £250, unless you buy a DX from Syco in which case you get it free). This allows the DX to work in omni mode (transmit and receive on all channels), and it will then control keyboards like the 106 and many others. It also doubles the DX memory.

It worked — the 106 changed patches accordingly when you punched a new sound into the DX. Pitchbend and modulation also controlled both keyboards. The 106 would play but would not control the DX (the latter's patches would not change when a new sound was selected on the 106). This could be seen as an imperfection of the system, but then it is more likely that the DX would be chosen as the controlling keyboard. Where this system was really impressive was in use with the drum machine, the 8-track sequencer, and Six Traks keyboard. Through the MIDI linkup even patch changes could be programmed into the sequencer.

From there we moved forwards and backwards in discussion. To the future: the modules. From the Roland system we tried a JX3P module (with built-in dynamics) and the controller keyboard (with piano-weighted action). This is for keyboard players who don't want to look like Rick Wakeman on stage, Kendall pointed out. You can have all these rackmounted modules (each a complete synth) tucked away at the back of the stage and just this one grand ivory-toothed beast to call them all up together, split (keyboard), or dual. Great idea but... expensive. Each module might set you back £800 or so. And, of course, the keyboard is £1600 and doesn't make a sound on its own. But no doubt prices will come down, and in such a system MIDI is the key.

Moving backwards: what about my old keyboards with their medieval CV or DCB or whatever triggering devices? Are they totally redundant now unless I can summon the energy to ever play in real time again?

No. There are ways to link most things, says my guide. There's the Syco Systems AMI (£199), a microprocessor-based device which converts analogue into MIDI codes. But it means you can drive MIDI synths with analogue synths — not the other way. But at least we can plug our dusty old pre-MIDI stuff into something apart from the wall socket.

So MIDI does work... conditionally. First condition: you must decide what you want to do and ask the right questions (preferably put to someone who knows the answers). Second condition: see and hear it working. Unless you're really big on binary code, mere theory may not be relevant.

Kendall says MIDI will get better — it has had teething problems — but it's up to users to ensure that things happen. Get angry and complain a lot!

Most of the problems encountered in MIDI (dis)connections could be cured if a few lines of the appropriate software were changed. It is apparently only a matter of them knowing what we want.

But back to the point of getting-the-most-for-our-money. It is more important (with MIDI) than ever before to plan your system. Random buying will almost certainly dump you and your ravaged bank balance in the MIDI maze.

In the middle of this keyboard victory roll, you just happen to be one of the remaining half a dozen people in the British Isles still playing electric guitar... at least that's how it seems. So what can this MIDI stuff do for you?

Thought number one is the Roland GR-700, the latest in that company's line of guitar synths. Much has been written about the controller, the 64 memory programmer and the synth sound section itself (very close to Roland's JX3P keyboard).

But there is a MIDI out socket at the back of the 700. The pitch to voltage conversion that makes sense of the guitar's string vibrations is done within the 700 so anything that leaves via the MIDI sockets can be understood by any MIDI-ed synth. The silver, wedge-like, G-707 guitar could control them all.

Perhaps one of the most successful combinations is with Yamaha's DX7, not because of any additional facilities it might offer but because of the nature of its sound — it's especially good at clean, natural percussive and plucked string voices and, coincidentally enough, so are guitars. I've heard several guitarists say they feel more at home with the DX noises. The behaviour of a twanged string, with its sharp, initial envelope and changing harmonics, runs some close parallels with the interaction of Yamaha's sine wave operators, the technique behind FM Synthesis. Of course, if Roland hadn't done the pitch-to-voltage conversion in the first place, you'd never be able to hear the results.

This year's NAMM show in America brought rumours of companies working on freestanding MIDI pitch-to-voltage converters. That way you wouldn't be paying for the JX3P built into the GR-700. You might not need it, nor like it. You might already have other suitable synths of your own. But as yet no-one's come up with any firm products. The most obvious contenders must be Roland themselves. They'd have the least work to do and it would fit with their current modular philosophy.

But let's go with what we've got.

DX7s, unless doctored, receive on MIDI channel one and that's how the GR-700 sends. Pitch and dynamic information are transmitted fine, but we're soon into the anomalies of analogue-based technology attempting to control a digital synth.

You can tune the G-707's strings with the machine heads, or bend them for pitch vibrato, but the Yamaha won't listen to such fine detail. It deals in semitone increases. That's why when you keep bending a string, nothing will happen for a while until suddenly the pitch leaps to the next fret. Same goes for using the tremolo arm. The GR-700 itself offers you the choice between normal, smooth pitch bend and this chromatic approach. MIDI cannot. (The strange side effect is that your guitar can actually drift out by a quarter tone on every string yet the Yamaha will stay in perfect tune.)

Both Roland and Yamaha will react to dynamic information — the harder you hit the string, the louder the note — but they behave in different ways. DX7 sounds travel to further extremes, they can be quieter and louder than the lower and upper limits of the Roland. But the GR-700 operates more evenly — closer to how you'd expect a real string to perform.

Linked to the DX, only a slight increase in plectrum pressure is needed to produce a considerable increase in volume. An advantage here is that with both synths connected, you can make the GR-700 or DX7 dominant by judicious twanging, without volume pedals.

Again, the way the Yamaha responds to dynamic information is truer to a genuine guitar — the brightness, tone and harmonics all change, not just the volume.

Now we come to the bits that won't work and this is where MIDI begins to flounder. You can't change patches on the DX7 using the GR-700 footswitches, so not surprisingly, you can't edit any of the Yamaha sounds by applying the GR-700's PG-200 programmer. In other words, the DX has to be close to hand.

You can still turn strings off and tune them electronically to produce open tuning chords, even though the guitar itself stays in the normal E/A/D/G/B/E mode. (One of Roland's demonstrators used this facility for a cunning ploy at a recent music fair. He set all eight slots in one memory bank to the same sound, but gave each a different tuning. By keeping the Roland's hold pedal down and changing memories, the synth produced a complete chord sequence for him. We tried to copy this on the DX but fell flat because the hold switch won't work with the Yamaha. Shame.)

What of the G-707 guitar itself? Shortfalls here, as well. The volume control works, but the extra knobs included for changing the filter can't function for the DX because there's no filter to talk to. One interesting spin-off of the GR-DX combination is that for the way I play, at least, there seemed to be fewer glitches. Or perhaps it's just that digital sound creation techniques won't show them up so boldly.

So the MIDI message for guitarists is even plainer than for keyboard players. Yes, it will get you started but you still cannot mix manufacturers and expect them to communicate like long lost brothers. If MIDI is a language then guitar synths have learnt yes and no and keyboards are up to short sentences, but we're still a long way from decent conversation.