Reversed Speech Intelligence Gathering


Tools, Techniques, Theory, Examples

I.  A Different Model For Reversed Speech

In all fields of art and science there are different "schools of thought" on how the subject matter should be taught, organized, and practiced.  There are often levels of expertise, described by terms such as basic, advanced, graduate, master, etc.  And there are usually schools that focus on a certain area of specialization, attempting to refine the methods and techniques most suited for a particular purpose.  The study and application of reversed speech (RS) is no different from any other field in this regard.  Being a field so rich in potential benefits, RS will expand significantly as an increasing number of sharp, creative minds are drawn to its study.

The techniques and results featured on are a form of RS specialization; essentially an intelligence gathering method designed to discover and reveal plans and events occurring behind the scenes in any human enterprise.  As such, it has a specific methodology all its own, involving tools of significant precision, techniques to enhance clarity, and a structure of well thought-out rules and logic to enhance and measure the reliability and usefulness of the information it produces.  Because of this it is exceptionally good at what it does.  On the other hand, this form of RS might not be suited at all for other uses, such as psychotherapy or determining how to relate to another person face to face.

If you have studied a basic or introductory level of RS you were given certain rules to guide your activity.  These rules are designed to help you achieve basic competency without getting into too much trouble, and also to protect the "image" of RS from those who have not yet reached a sufficient understanding to branch out on their own.  If you are interested in RS-INTEL you should understand up front that according to the most commonly distributed rule set, intelligence gathering via RS is impossible and forbidden.  What a human tragedy it would be if that were actually true!

By definition, any form of intelligence work with RS is directed at information that is hidden, and consequently unspoken.  However a common rule for students requires that "valid" reversals must relate to the subject matter of the "forward" speech.  While it can easily be shown that a single word in forward speech can create a thought-form that can then trigger an apparently unrelated reversal, this is certainly not happening in every case.  The unconscious mind is far more active and complex than previously imagined, and RS proves this virtually every day.  Still, within schools of RS that depend on the relatedness rule to keep life as simple as possible, RS-INTEL and any similar endeavor is by definition "not valid", and can be declared so without regard to how such a statement might be perceived by others.

Luckily enough, there are no RS police and advanced practitioners are perfectly free to experiment, discuss, evaluate, and create specific formulas in the endeavor to achieve more useful results.  RS-INTEL, for example, relies on a fairly extensive set of considerations for the evaluation of reversals.  Rather than "validity", these rules are designed to assess and assign levels of "confidence" to the operator's interpretation of what has been found.  Generally, any reversal that is understandable is considered valid, unless another practitioner suggests otherwise and offers a good technical reason.  For instance a short phrase might turn out to be a part of a longer phrase that was missed.  However, a valid reversal does not necessarily have any useful meaning.

Because of this, the more important effort is to identify the meaning, significance, and value of the words that were found.  This is a much more appropriate concept for the intelligence field, which deals with degrees of probability in an environment where any scrap of information can prove vital - or fatal.

Considerations that effect the confidence level assigned to a specific reversal or cluster of reversals include the following questions.  Are there enough words to complete a phrase or sentence, and does it include a subject and verb?  How many alternative interpretations might the phrase suggest?  Does more than one reversal refer to the same subject?  Do they each add additional detail?  Are related reversals close to each other, ie. within 2 or 3 seconds?  If not, is there a trigger word (or concept) within the forward speech that might account for the multiple occurrences, and is the trigger word itself significant?  What are the indications of emotional overloading, and are they consistent with the interpretation?

Do the reversals relate to a known issue (from some other source), and is it likely that the person who is speaking is aware of or involved in that issue?  Do the names of any people, programs or objects known to be involved in the situation also appear in the reversal(s)?  And of course, the Gold Standard for confidence, is another individual also providing reversals that confirm the existence and details of the same hidden topic?  Do they add additional detail and is the overall picture self-consistent?  If there are differences is there a reasonable explanation, eg. caused by differences in rank, security clearance, or personal/cultural viewpoint?

An RS-INTEL operator is greatly assisted by working in tandem with other more traditional specialists who can better identify the many personnel and program names, abbreviations and occupational jargon that appear in RS, and who knows more about any history involving the organization or program being investigated.  This can cause the value of the end product to double or even triple.  Likewise, the end product is more valuable when two or more RS-INTEL operators work together and can review and contribute to each other's logic.

At this time there is no textbook for RS-INTEL, primarily because it is a work in progress.  For instance, a formula was devised to assign numerical values to the various factors effecting "confidence".  If successful, a fixed formula would allow a final computation, ranging from zero (unknown) to one hundred (solid truth), by the use of a simple computer program.  However, when applied to all the factors that went into the Easter Nuke investigation, the formula produced numbers in the range of 400 to 500% and it is not clear how the math can be scaled back without degrading it's performance in less substantiated situations.

For every high-confidence interpretation, there are hundreds that do not make the grade, and a similar horde of reversals whose meaning or significance is totally unknown.  It is important, however, that even these must be preserved in some sort of searchable data base.  This has increased in importance today, as we find that six and seven year old reversals that made no sense when they were first found, now link up with current situations which are portrayed in the media as coincidental or unplanned.

Before you ask, no one I know of is currently interested in becoming an RS-INTEL guru, or teaching it in any formal way.  In the past it has been communicated and often greatly enhanced, when people join together for specific projects.  Those who are good at it generally keep low profiles.  You've probably never heard of them.  Consequently, it is a wide open field characterized more by private discussion aimed at increasing accuracy than anything else, and everyone is welcome to explore it for themselves and contribute whatever they can.

    -- STRIDER

NOTE: The source files put up on a file sharing service have apparently expired.  An attempt to reconstruct these will be made, but at the moment there is simply no time to do so.

How RS is presented on K-W.Com

When Ken-Welch.Com went live it had only one purpose, which was to stop the Easter Surprise by publicizing the Bush Administration's audio fingerprints that were all over the plot.  Because the article used reversed speech, many RS students visited the site to see how RS was being used here.  In a school setting, RS students are accustomed to having all the source material (audio files) laid out for them so they can practice reproducing all the reversals that are presented.  However, Ken-Welch.Com is not a school, and RS students are not it's intended audience.

Regardless, a number of students became confused when not all the "forward" source material was presented.

To put it bluntly, the site was already pumping out five megabytes of audio to each visitor who wanted to hear the whole story of a loose Nuke in Galveston Bay, and simply could not afford to provide more.  But more than that, the vast majority of people receiving reversed speech intelligence material, in whatever setting it is provided, do not work with RS themselves.

Consequently, the policy for the reports on the site is to limit the amount of forward speech provided.  The rule for RS reports on this site is to provide one really clear example of forward speech being reversed and then isolated - for each sound clip.  This will always be enough for someone to experiment with themselves and validate the reversal process if they want to - and also enough that a visitor can be fully satisfied that it is indeed the person who was named, and the particular event or presentation that is listed.

Additional snippets of forward speech are provided primarily to assure the listener that it is the same person speaking, and give some indication of where they are in their speech and what subject matter may have prompted the reversals that follow.  These will also be reversible, but may require advanced skills.  When not needed for the purposes given above, additional forward speech may be omitted entirely since the majority of listeners already have a good grasp of the concept and simply don't have any use for it.

That said, I (KW) also have considerable sympathy for RS students, both independent self-starters like myself, and those in proprietary programs.  I have an interest in helping them learn their craft when I can.  Some of them might very well join us in attempting to change the world for the better.  I feel students are best served by challenging themselves with the original material if it is still available.  Knowing that the reversals are in there because you've already heard them from another source is a very strong incentive to stretch your skills.  But students can also benefit by access to the actual working files (.wav) that were used - especially if those files contain a marking system that shows the reversals in place, with labels and selectable start and stop positions, and ours do this automatically.

Consequently I have always been willing to provide the working files for any report segment to anyone who 1) wanted to check the technical aspects of our work, or 2) wanted to learn advanced techniques by figuratively looking over our shoulders as we isolated the reversals that supplied all that terrific information.  Oddly enough, even though there was apparently some controversy in other circles over the material presented here, no one ever asked.  I don't know why not.

Nonetheless, and without being asked, I've decided that the site should provide some good source material, and in some cases tutorials or notes, for RS students who really want to explore our techniques and methods.  Remember that this is not the purpose of the site, so it will be done strictly on an available-time basis - and there's not a lot of time available these days.

All the public audio that was presented in the nuke report is available in the mirror file.  This is a file that was offered to other websites as a back-up if K-W.Com disappeared.  It's a perfect copy of the site on it's first day of operation, which means it has a bunch of folders with web pages, graphics, flash players, and MP3 files.  If you download and unzip the file, you can browse the folders and play the mp3's directly.  They contain no markings or labels, but having the actual sound clips will give you better sound than trying to record off the web while they are playing.  Send a blank e-mail to  "mirror"  at - and put the word "student" on the subject line.  An autoresponder will send you the current location of the file at a free download service.  It is a 3.5 megabyte file that unzips to 5 megs.

All the working (.wav) files are destined to be put on a CD, that can be snail mailed.  This requires a lot of organizing at this end and will take time.  It will be a fabulous resource for RS students willing to use the same software we do - because all the reversals will then show up, marked and selectable.  It will NOT be free, but should not exceed the cost of an average textbook either.

I will also pick out what I think is the best segment from the nuke report, and put the supporting files on a download service as we did with the mirror file.  That will be entirely free and you will then have something that you can wow your own students with someday.  I've already done this with the twenty-times-smaller Bush On Cocaine file.  See the tutorial below.  -- Ken

Source Files and Notes - Bush On Cocaine

Currently the most recent RS-based article on, the Bush On Cocaine story is an excellent example of using RS to ferret out something that is otherwise hidden.

I was surprised when someone I previously had great respect for dismissed this article as an "elementary school claim".  I still have not figured out exactly what this implies.  It seems to me the if the President of the United States has appeared for at least one White House function stoned on Cocaine it should be significant, despite the fact that we live in very strange times.  However I finally understood the person I mention had not actually read the report or listened to the reversals, a common situation that is the bane of writers the world over.

The reversed speech in the Bush story is a good example of how reversals from a single recording can confirm each other, and supply enough information to create a good degree of confidence in the resulting interpretation.  Two of the statements are complete sentences and taken together do not leave much room for alternative explanations.  External events, in this case the distinct and unusual mood of the speaker, also add confidence.

The unconscious statements found by reversing the President's initial remarks are:

  I'm all goofed up.
  I'm on powder.

I have assigned a confidence level of 90% to my conclusion that the President was high on cocaine.  What about the other ten percent?  This stems from the fact that cocaine itself is not actually named, and therefore there is a possibility that some other drug might habitually be referred to as powder.  Junkies invariably have a pet name for the drug they use, because the actual name carries so much negative emotional freight.  If it could be shown that some other drug is commonly called "powder" by it's users, then the cocaine hypothesis would be open for debate.

RS students who want to examine the sound clip included in the report can download the MP3 file directly from the website.  Right click on the link below and use your browser's "save as" command to download the file to your own machine, or drag the link to your downloader if you use a separate program for this purpose: Link to MP3 File.

This will give you the finalized version with greater fidelity than trying to re-record it while playing it off the website.

For tutorial purposes, the working file that was used to produce the MP3 clip has been uploaded to a free download service.  It is a mono .wav file recorded at 44k bits per second.  It is not "zipped" and occupies 4.3 megs.  Click on the following link to open a new window at the download site:


The picture above shows the full graphic representation of the sound in the source file.  This is a screen shot from my audio editor, Sound Forge.  If you are using one of the programs in the Sound Forge family, you can quickly zoom in on any segment of the file and see how the operator (me) marked specific segments, typing the words he found into the labels that appear over the segments.  In the nine years I've worked with RS, I've found this feature to be the most valuable tool of all.

In fact, I was so impressed with the program when I first found it, that in late 1998 I became the first person to offer a software solution for doing reversed speech, shipping boxed versions of the program that I purchased from it's originators, Sonic Foundry, with a book describing how to adapt it's features for RS, how to find many more reversals than other methods were producing, the discovery of overlapping reversals, and so on.  Unfortunately this was not a great marketing success.  I think I sold about 16 copies.

As it evolved, Sound Forge developed into an exceptionally powerful professional version (big bucks) and a reasonably priced "home version" that is perfectly good for RS.  The home version normally carries the word "studio" in its name.  The current version (now owned by Sony) is called Sound Forge Audio Studio.  Happily enough, these programs have always been offered as try-before-you-buy shareware.  So you can download the program from Sony for free.  Naturally it will only work for so long and then you must purchase it if you wish to continue using it.  Current versions are at least two upgrades newer than the one I use, and I'm sure they have all sorts of new bells and whistles.  If you need to get the most bang for your free buck, and are only downloading for the material presented here, you might hold off on getting it until there are two tutorials here instead of just one.

If you've downloaded the .wav file, and can open it with a program that shows the Sound Forge notations, you will see that there are three basic parts to the file.  A graphic of the entire file is shown at the top of this tutorial.  The picture below shows the first part, which is the full forward speech from the first 29 seconds of the President's remarks.

I've marked the two areas that contain the cocaine reversals, designating them as segments 1 and 2.  If you are using a different sound editor, you can still pick these out by the shape of the wave forms shown in your editor.  If you are at the same "zoom" level they should look just the same as pictured above.

As you move to the right (forward in the file), you will find that the next portion of audio is the "segment one" material all by itself.

In Sound Forge, you designate beginning and end points for a particular piece of sound and mark it as a "region", which allows you to type in a description that you can see on the screen any time you open the file.  You can have regions within regions, and regions that overlap each other.  Here you see that I've set up an area and selected it (that's why it's pink) that contains the forward version of the segment 1 sound, and then the same sound reversed and stretched out.  The region label contains my notes that indicate the reversal was stretched to 138% of it's original length. That means that when I play the stretched sound at standard speed it will come out much slower.

Time stretching allows you to slow down the sound without having the recorded voice drop in pitch.  I believe this helps greatly in allowing the ear to understand speech reversals.  Look at the picture above and see how the sound on the left has been pasted to the right, then flipped left-to-right to make it backwards, and finally stretched out to so it will play slower.

Time stretching formulas vary, and some are more suited for the human voice than others.  If you have choices, try them all before selecting one.

On the screen, regions are shown with vertical dashed lines at their boundaries.  Within the Sound Forge program you can double click between the lines and just that segment is selected, allowing you to play only that sound - all by itself.  Note that I have marked the two reversals that appear here, typing the words into the region label, so that I don't have to look for them again.  These are "Powder!" and "I'm all goofed up".

If you have the source file open in Sound Forge at this time, you can select either one of these by double clicking on them, and hear the reversal by pressing the space bar.  Note that if you just play the whole segment as it is, it is much more difficult to pick out the reversals as they race by in the midst of other backwards sound.  In fact, if you listen to more than about five seconds of ordinary backward speech, your mind will begin to deliberately tune it out because it makes no sense.  Consequently, when a good reversal comes along you may not even notice it.

Just a bit further to the right, the two reversals have been copied and pasted by themselves which again allows you to hear them better:

Notice the change in sequence.  In reversed sound the flow of time is also backwards so, even though your program is playing the sound from left to right, the flow of time is right to left.  In the picture above you can see that I've returned the reversals to their original sequence.  Experience suggests that when two or three reversals come out almost on top of each other, the actual sequence may not be relevant.  Therefore, the order in which you present them to others will sometimes be a judgment call involving issues of clarity or understanding.

Finally, in the last section of the source file you will see I've done exactly the same thing with "segment 2".

In the reversed and stretched portion, I've marked the reversal, "I'm on powder" as it occurs within the backward sound.  People who are hearing reversed speech for the first time often wonder if it is possible for the RS operator to insert sound for his own purposes.  They don't realize that anyone who wants to can simply reverse the sound again and play it in the normal direction.  When that happens, anything that had been inserted would stick out like a sore thumb.

Note that when I lifted out the "I'm on powder" reversal to use it in the website sound clip, I increased the volume of the first word about 5db.  There are three words here, which are the three "bumps" in the sound form.  The first word, I'm, was spoken at pretty low volume.  You can see that it has been boosted so it is now just as "tall" as the word next to it.  This will make it much easier to hear, particularly the first time it is played.

The information in a reversal (actual words), and the volume at which it appears, are not related.  You can't change the words by changing the volume, and in fact the original volume, like the original pitch, is created by the forward speech.  The reversed speech is simply stuck with whatever is coming out of the mouth at the time.  In practical terms, a segment of reversed sound that contains large differences in volume (high points vs. low points), otherwise called "dynamic range" is much harder to work with because loud words tend to mask the softer words next to them.  A good tip for any RS researcher is to keep an eye on the dynamic range of original material, and if it is too high simply to reduce it across the board if you have a processor or editor feature that will allow you to do so.  Mostly you want to bring the peaks down, and if you can raise the level of the valleys a bit as well, then you will find the reversals much more easy to understand.  Don't be afraid to do this by hand, syllable by syllable, if there is no other way to "bring out" a reversal.  After all, once you can hear it properly it might turn out to be different from what you initially suspected.

That's it for now.  Download the files even if you don't have Sound Forge, and have some fun with the source file.  Practice locating the reversals, and adjusting the right and left boundaries until they are perfect; a very critical skill.  See how you might cut and paste the sound to create an effective presentation for others who are not used to listening to RS.  And by all means save that compressed MP3 file someplace where you won't lose it.  Years from now you will be finding opportunities to say to someone, "Say, did you ever hear the one about Bush on drugs...?"

    -- Ken Welch, Houston