Analysis of Ken Welch Documented Reverse Speech

By Pierre Alroy

Pierre has a Master of Arts in Applied Linguistics, and has studied phonetics and phonology at both Macquarie University and The University of New England. He is also a certified Reverse Speech Analyst. He can be contact by emailing roipierre@hotmail.com
 

Introduction: In March 2006 Houston researcher, Ken Welch, made an elaborate claim on his website that he had found speech reversals on key government officials indicating that Houston was going to be nuked over Easter. This was part of secret Government plot to create a second 911 for whatever reasons of their own - essentially the US government was planning to nuke its own citizens. See his website here.

We immediately went to his website and analysed his alleged reversals. We were dismayed with what we found. Most of his alleged reversals did not exist, some were altered from the original sound tracks with sections of sound cut out from the tracks to make them sound like something else. We began to receive emails from concerned people wanting to know our opinion on Ken Welch's claims, so we issued a disclaimer about this event and Ken Welch's reversals - click here to see this release.

Not surprisingly, Easter came and went with no nuclear attack, as we predicted, and we thought this was the end of the matter. Not so. Incredibly, Ken Welch then claimed that he had prevented the Nuke attack on Houston by warning the public beforehand and thus had thwarted the diabolical plans of the US government. Of course, this claim completely misses the most obvious explanation - that he was either wrong about his original reversals or he had misinterpreted his findings, or both.

Furthermore, he then added a page to his website purporting to teach others how to do Reverse Speech (which he conveniently calls reversed speech.) This is despite the fact that he has no formal training in the field of Reverse Speech, his documented results are not of sufficient quality to be set up as a standard for others to follow and, despite his attempt to spin this event in a different way, he now has one dismal failure in his track record.

And so, with no choice, we have continued with our analysis of Ken Welch's alleged reversals with a detailed professional linguistic and auditory analysis of his posted results. Our thanks go out to Reverse Speech Analyst, Pierre Alroy, for this report.

David J. Oates
backwards@reversespeech.com

Pierre's Report

BUSH

LINGUISTIC ANALYSIS

Easter: The documenter has taken the sounds off the end of a long section of sounds. It is not representative of the word Easter and is projection.

Prime it: There is a rounded mid low central vowel followed by a labialised /r/. /n/ occurs rather than /m/ and there is a shortened high dipthong vowel as the final. Prime it is not represented.

And you fire it: fire it has a /p/ initial with some aspiration. The /r/ has extra voicing. The sounds resemble ‘pirate’. And you is not indicated by the sounds.

In the ship: The documenter has cut this out of a longer tonal unit. He also sliced of the body of /n/ after its onset in an attempt to produce a /p/ at the end.

I’ll use it: The documenter has cut off the initial sibilant and the first part of the vowel (which are an integral part of the sound segment). If you go back to the longer forward speech section earlier and reverse it, the sounds are fairly representative of sell use it. This sound segment follows what is documented as Nasa. However, there is a shortened higher than normal mid-low back rounded vowel sound after /l/ and before the onset of the sounds representing use which is representative enough of or uttered in this phrase in normal running speech. This can be hard to hear probably due to the length of the /l/ from onset to closure, but can be discovered on closer listening and becomes clearer when slowing the audio down. Note that this short vowel sound is not I as the movement of the tongue up to the vowel is use may give some the impression of it. It should also be noted that I’ll is not there as was documented. Cutting sell in half where the tongue begins to rise will get I’ll.

Near Nasa: Nasa is represented by the sounds. However, the documenter has cut out a phoneme /d/ after the /r/ phoneme, immediately before Nasa. The sound segment before Nasa can be perceived as two syllables rather than one with the second syllable approximating ‘eard. This section is somewhat like “’eard Nasa sell or use it.”

Place to go for Easter: Easter is fairly representative. The first four words are not representative of the sounds. Some similar phonemes have only given a wrongful impression of it.  There is no /l/ in the first documented word, and there is an /r/ in the third documented word. For is unlike the sounds. There is a velar that is articulated more forward in the palatal region. Lucinda unicode does not have the phonetic symbol for it so I have use a palatised /g/. When this happens, it can sometimes give the impression that the sound is a /d/.

Nuke – place to go: The documenter has cut off a sound segment behind the forward words to this in an attempt to produce the word nuke. The sounds left in place, however, are very unlike the word. I am amazed at the lengths the documenters have been prepared to go to create target words.

 You are all slaves: The documenter has cut the section out of a longer tonal unit. The sounds are resonably phonemically similar to slaves. /v/ is more like the labiodental approximant /ʋ/. The first three words documented are not represented by the sounds.

SUMMARY

Poor listening skills and ability has resulted in a majority of sound segments incorrectly documented. The documenter heard a fair amount of I’ll use it and near NASA correctly, however, ended up cutting out important sounds to achieve what he wanted. One of the Easter documentations is quite close, the other is very far from it. The rest of the documenter’s findings possess no value.

GENERAL PACE

LINGUISTIC ANALYSIS

Nasty purpose: On the surface it has some similarity to the documented words. However, in nasty there isn’t a /t/ after the sibilant /s/. The nasalisation of the vowel sound /æ/ could give a false impression of this happening. The final vowel sound is also too dissimilar to the final vowel sound in nasty. The second /p/ in purpose is loosely formed and is more like /w/in that the lips are not quite pressed together and there is more roundedness. However, in normal speech the second /p/ can often be pronounced with less lip contact producing a more approximant sound. Some people will also round their lips at this point. Purpose is probably within reason phonemically.

To be with God: The sounds can give the impression of the phrase. Like vowels that would normally be found in to be with, the sounds are also high vowels located from central to frontal positions. The vowel articulations are not comfortable with the usual positions however they are ‘close enough’ to give the impression of the documented phrase. There is not enough evidence of a /t/ articulation at the start. There is a /β/ that resembles /b/ and perhaps a labiodental approximate /ʋ/ resembling the documented /w/ in with, however there is no ending typical of the word. “God” doesn’t have a consonant ending and ends with a glide to a higher vowel.

They tried to kill me – dead: The sounds are fairly representative of the words  ’n they tried to kill me.  Even though the articulations normally would remain in the front of the oral cavity, it seems that enough force occurred in the velar region to approximate a /k/. The last word documented dead does not occur and only a vowel sound is evident.

Why Easter: There is phonemic similarity to the documented words. Why has a vowel sound that begins in a low back position and ends around a mid-high front position like an Australian accent. However, the sibilant in Easter is quite distorted. There is also a sibilant-like noise at the end of the documented word. 

The bomb - made in Russia: When listening to the documented word bomb it can give the impression of a plosive /b/as the initial, however on closer listening one should be able to hear a voiced sibilant that is labialised (rounded). A plosive sound is produced by the speaker’s vocal force on the microphone at that point. At the end of the documented word bomb there is an /n/ + vowel sound. The initial in made is not /m/ but /n/. There is no /d/ ending. In in Russia a vowel diphthong as in ‘show’occurs at the end but much shortened. Apart from this, the sounds reflect it. Nevertheless, in Russia exists in isolation and without other word-like examples around it, it is unremarkable.

Who is he gonna kill: The only documented word with adequate similarity is kill, however there is the addition of velar fricative noise. The first four documented words are represented by the sounds.

Ship – sucks – Houston – bomb: The first two documented words are taken from a slightly longer sound unit (see table above). The documenter has cut what he believes to be sucks from it, then to create ship, he cut slightly before the onset of what is documented as sucks and cut at the end of the vowel sound in the same documented word in an effort to produce a /p/ sound. It seems that the documenter is prepared to go well beyond any logical or reasonable bounds to produce a desired word. Houston is part of a longer sound unit that is gibberish. It is far fetched to claim that it is a real word. The documented word bomb is also part of that gibberish. It doesn’t exist.

He threatened you yesterday: There is a phonemic similarity to threatened, however there may not be  –ed . In you the initial is a velar nasal with a short /ɪ/ followed by another shortened vowel sound with a rounded high front articulation. The nasal + /ɪ/ may give some the impression of ‘me’, or the movement from the first vowel to the second vowel may give others the impression of ‘you’.  The documented word yesterday is not represented by the sounds. The sounds are quite similar to desperate, except that the there is a palatal plosive (which gives the impression of a /d/). The palatal plosive font is not available in Lucinda unicode, so I have used a /g/ with the palatal diacritic to show the tongue is more forward than a velar articulation.

That means nuthin’: Clear projection into the sounds by the documenter. The speed of the sounds where means is documented combined with the fact that there is a possible bilabial nasal and sibilant has given the impression of the documented word, however in reality it is too dissimilar. Nuthin’ is articulated as nufnik.

Now we are the terrorists: Now we is indicated by the sounds. There is a voiced sibilant between the documented words are the taking the place of th in the. There is also a /w/ between we are. Terrorists is not there with the sounds approximating ‘serosit’. There is nothing documentable as language with only sounds representative of now and we.

SUMMARY

The majority is projection. However, the sounds are fairly representative of the words they tried to kill me (but not dead). Why Easter has a fair amount of phonemic similarity with the sounds with distortion mainly in the sibilant(s).  A section of one of the findings, in Russia, possesses some phonemic similarity, however the ending is incongruent. Another small section of a documented reversal, threatened, is phonemically there. The documenter has carried out a rather strange action on ship – sucks as set out in the analysis.

BERNANKE

LINGUISTIC ANALYSIS

Soviet nuke: The documenter has cut this from a longer tonal unit. There is an /f/ initial (this is most evident listening to the longer section). There is no articulation to a /t/ in soviet. It flows straight to /s/ (first vowel /e/ before /s/ is higher than normal but not indicated through lack of appropriate diacritic). This is followed in his reversal section with /n/. Soviet is the result of poor listening skills and as for nuke, where is it?

By the pier: The initial in by is /ð/ (th), the initial in the is/b/ and the initial in pier is /s/. There is nothing documentable.

Hidden inside: I have taken the transcription from the reverse of the actual forward section, not from the Welch reversal which has been doctored. There is a single /n/ not two separate ones. The documenter has cut the single /n/ into two to create the desired ending for the first word then separated the words. He has created an extra syllable that wasn’t there. Also in the ‘real’ section there is a short sibilant initial followed by /k/. The /s/ in inside documented by Welch is more like /θ/ but perhaps with an /s/ ‘colouring’. The final sound (which is most evident in the longer section) is a very soft /θ/. There is nothing documentable here.

The Russian’s nuke: Russians possesses phonemic similarity. The vowel is more forward than usual. There is no evidence that the initial of the documented nuke is there. It is a plosive not a nasal.

Near Beaumont: The sounds come out as Nikolai. This can be heard fully behind the longer uncut section. The documenter has sliced the ending of it. In the longer section this is followed by another short sound segment which is clearly part of it and has the character of a surname. This is then followed by what appears to be Russian above.

In the ship: This is part of a longer tonal unit. A couple of minor incongruencies, but a serious one is that the ending is clearly /f/. This is not the documented phrase.

Fearsome – fiercely evil: For the section documented as fearsome there is a /b/ initial, following the /r/ there appears to be a /k/ with a more forward palatal articulation followed by a short articulated /n/. There is no sibilant as documented. After the /m/ there is another vowel. The segment is unlike the word documented. The section documented as fiercely evil – the section has an /s/ initial. There is an /n/ where the documenter put c and there is no /l/. Evil possesses a devoiced /b/, a vowel that is much further and higher than acceptable for the word, and it ends with an alveolar not an approximant. The documenter cut it just before or at the onset of an /s/ leaving a /t/. The ending of the tonal unit is here, and it really depends on whether the end of the section is at the gap between the /s/ and /b/ in the forward speech where a /t/ can be perceived in the Reverse Speech or at the /s/ itself.

Clear out Houston: The initial is probably closest to a /t/ with mild aspiration. The sounds documented as Houston are unlike the word. The initial is /g/, there is no /t/ as the initial of the second syllable, and it does not end with an /n/.

Sended it in the ship: There is similarity with sended it in the, however, there is no –ed ending. It is more similar to send us it in the spoken in natural speech. There is a glottal t /ʔ/ in it. Ship is not there. There is an /l/ followed by a clear /f/ which the documenter cut off in order to create a /p/.

SUMMARY

The only interest here may be the existence of Nikolai and Russian. The rest of the documented material is of no value. The documenter once again has been slicing off integral parts of sound segments to try and achieve the desired word, but this time he has even doctored a reversal by adding an extra syllable.

RICE & SCHWAB

LINGUISTIC ANALYSIS

April fear it: April is projection. The initial is a velarised /n/, there is a /w/ rather than an/r/, and a full vowel ending rather than an approximate /l/. The documented words fear it appear to be represented by the sounds.

Yes I know we are fearing it: The only words that could be represented by the sounds are I, know and we’re (not we are). The sounds after this approximate ‘siri you’.

Easter: This is clear projection. The documenter has clipped it out of a longer set of sounds. It is not representative of the word.

Kill – Bush made you kill: Instead of taking the reverse of the selected forward section which would have produced jaʊp̚ neɪgɪbluːəs kjæl the documenter has cut each word separately, reversed them individually then rejoined them to create the desired sentence. The sounds are not representative of the documented words anyway.

You’ll see their war now: The only documented words that could be represented by the sounds are you’ll see taking into account his German accent (Note that /ʋ/ is not /v/ but is a labiodental approximate possessing more of a vowel quality than a consonant). The rest of the documented words are simply not close enough to the sound segments.

She saw the nuke: Listening to the longer forward speech section, /ʧ/ occurs as the final sound of the tonal unit. However, the documenter cut it off to achieve the desired target word. In what was documented as nuke, the mid-high unrounded vowel is articulated higher so the tongue is closer to the palate and there is a puff of air. This palatal activity may incorrectly give the impression of a velar/k/ following. She saw could be represented by he sounds, however the rest is not representative of language. 

Not yet – Easter you know: Not yet is reasonably close to the sounds. There is no /t/ in Easter; a short glide between two high frontal vowels occurs initially, followed by /s/, followed by a high front rounded vowel with a short movement to a high front unrounded vowel slightly lower. You has a front rounded vowel rather than the back or back-mid rounded vowel typically produced. Know is produced with a vowel in the vicinity of a mid low back vowel, however, this is not so unusual due to the accent of the speaker. However, the tonal unit continues on at this point to produce further sounds. The documenter appears to have cut it at a point to produce the desired words. Easter you know does not exist.

On the ship: The documenter has cut into a (longer) tonal unit to create the desired sounds. The labial /w/ has been edited at the beginning to create on, and the documented material has been cut at the onset of a voiced labial approximant /w/ in an effort to create a labial /p/. Sounds do not represent documented words.

SUMMARY

I am stepping out of my intended purpose to just produce a linguistic analysis, but what the documenter has done leaves me speechless! It is quite clear that the documenter doesn’t know what he is doing. The only thing he seems clear about is finding something that will support his fixation to produce certain target words. Words like April, Easter, nuke, Bush, and ship are projections. He has also cut pieces out and even rearranged sound segments in order to produce a desired sentence.

NEGROPONTE

LINGUISTIC ANALYSIS

(Glossary of linguistic words at end)

The nuke’s here: The initial sound is a distinct sibilant. However, the voiced fricative nature of the beginning can give the impression of the word the. In the sound section documented as nukes there is an /m/ articulation (where the /m/ occurs in forward speech) at the point of where /n/ should be. However, there is a possible /n/ occurring before the /m/.  Following the vowel sound an alveolar articulation occurs as a result of the ‘st’ in ‘extremist’ which is then followed by a velar articulation as a result of ‘x’ in ‘extremist’. The sounds for here have some similarity. There is minimal /h/articulation. The ‘breathy’ sound prior to the onset may give an erroneous impression of one. However, minimalisation or h-deletion is not an issue as it happens in normal speech. There is a palatal articulation at that point mimicking a high vowel or /j/. There then appears to be a labialisation of the /r/ giving it a rounded quality.   

Don’t exist: The syllable and intonation pattern, as well as some phonemes reflect the words don’t exist. However, the initial sound doesn’t indicate with any certainty the existence of /d/. A softened /d/ lies in place of /z/ and the final sound is a voiceless affricate. There is too great a phoneme dissimilarity and there is nothing documentable here.  

We’re registered evil: The articulation appears to represent an American we’re. However, the section documented as registered is not representative, nor is the section documented as evil. In regard to the second documented word, registered, a soft aspirated /t/ moving to a mildly voiced alveolar palatal fricative /ʑ/may have given the impression of the gis in registered. A high centralised unrounded vowel /ɨ/is found as the final vowel in place of the centralised /ə/ normally produced in the word. As to the third documented word, evil, there is an evident lateral articulation /l/ in place of /v/, a second vowel that is somewhat higher and further back than the schwa /ə/ typical for the word, and a mildly rolled /r/ as the final sound. The finding is not documentable as language.  

The ship: The sound articulations are reasonably acceptable for the formation of “The ship”. However, its isolation and lack of supportive information in both the reversed sound and the forward speech makes it unremarkable.

Houston: The documented sound reasonably articulates to Houston. However it is extracted from the centre of a continuous stream of sound of the same tonal quality and indicates coincidental sound only.

Will blow up: First word, will, is representative of the documented word. The sounds following begin with an obvious sibilant initial while the first vowel sound is highly unusual for blow, although can give an impression of it.  The sound grouping does not indicate documentable language.

Little mushroom: The sound section documented as mushroom possesses a retroflexed and labialised /l/ in place of /r/. The sounds documented as little are not representative enough. This is not documentable as English.

The nuke – right near lifeboat: The nuke is not there with the sounds approximating dethoop. Lifeboat is missing /f/ with a /h/ articulated in the velar region in its place. There is enough bilabial quality to the labiodental /v/ allowing a /b/ to be perceived. Near does not occur. Right may be phonemically acceptable bearing in mind possible t – deletion that can occur as a result of a following sound. The nuke is not there and the rest is not documentable as language.

There’s no accident: The sound articulation for no exists. The first word, there’s, lacks adequate similarity to the word documented, with the initial being a retroflexed nasal /ɳ/ and the final more postalveolar than alveolar. However, weaker words such as this are subject to altering affects to their phonetic structure in speech. The fact that the sibilant is further back is not a serious issue, but that the /n/ is retroflexed and the tongue is curled back to the soft palate to produce /ɳ/ is too great a distortion. The section documented as accident is not representative. It articulates as akshinet. The intonation and syllable structure is reflective of the documented sentence. Interestingly, the phonemic structure almost represents a drunken person slurring speech.

SUMMARY

There is really nothing in the Negroponte findings that can be confidently documented as language. Sounds reflecting function words exist, such as will, that, and no, however this is entirely unremarkable. Sound sections appear to be representative of “that ship”and “Houston”.  The vowel sound in “ship” possesses the high frontal vowel /ɪ/ but glides a little to a centralised /ə/ and “that” possesses a higher vowel than the standard, but both are within reasonable bounds. However, to hear these sounds in speech which is reversed is unremarkable, especially without the existence of other (complementary) language either connected to it or in its vicinity. With “Houston”, the sounds are pulled out of a larger long continuous sound field possessing a similar tonal quality throughout. One can cut a small section of a normal speech recording that is part of continuous speech and create a seemingly new word from it. The sounds for “The nuke’s here” possess similar characteristics to the documented words. However, there is a sibilant in the initial, and an /m/ evident. There are also other imprecise sounds. One must be careful. When we listen to any speech our brain seeks recognition of language from the sounds produced. In this way, it is very easy to “gloss over” unmatching phonemes in order to produce recognition of language. And, of course, mishearing occurs when we have an “investment” in hearing certain words. The remainder of the material documented is not acceptable as English.

GLOSSARY

Affricate: consonants that begin as plosives (t, d, k, g etc) and end as fricatives (see below for fricatives). There is a closure followed by a fricative noise. Typical consonant sounds are tch and dzh.

Alveolar: place of articulation on the alveolar ridge behind the upper teeth. Alveolar consonants include t, d, and n.

Aspiration: After some consonants, typically p, t, k, air is pushed out in the post-release phase producing a sound like h. An example would be the p in push.

Bilabial: both lips, for example b and p.

Fricative: manner of articulation whereby air is pushed out through a constricted passage giving a characteristic hissing or sibilant quality. Typical ones are f, v, s, sh, and th.

Labialisation: lip rounding of any sound normally articulated with the lips in a neutral or spread position (note: labial means upper and lower lips).

Labiodental: lower lip and upper teeth, for example f and v.

Lateral: A lateral consonant such as l in look, where air escapes along the sides of the tongue.

Monophthong/monophthongal: a single vowel sound where there is no tongue movement, for example the vowels in dog, cat, bit, seat, cup (compared to a diphthong where there is articulatory movement from one target vowel sound to another, for example the sounds produced by the vowels in boat, howl, main, like).

Palatal: the region on the smooth surface of the hard palate extending toward the soft palate (see velar). An example of a palatal is y as in yes.

Retroflexion: curling backof the tongue so that the underside of the tongue has contact with the alveolar-palatal region.

Rounded/unrounded: vowels can either be rounded or unrounded. For example, o and u, are produced with the lips rounded while a, i, and e, are produced with lips spread (unrounded).

Sibilant: main ones are s, z, sh, zh.

Velar: the soft palate near the nasal passage is the velum and consonants produced either by the back of the tongue coming into contact or into close contact are called velar consonants. Examples are k and g.

Voiced: vibration of vocal folds, typically v, g, d, and b.

Voiceless: lack of vibration of vocal folds, typically f, k, t, and p.