ESDA and the analysis of contested interview notes

Tom Davis

This paper was published in Forensic Linguistics 1 (1994): 71-89.

Introduction

In October 1987 I was asked to examine a set of documents. At that time I had been working part time as a forensic document analyst for thirteen years; this case, however, was different. It did not concern handwriting, as the large majority of my previous cases had, but the use of ESDA: the Electro-Static Detection Apparatus. I was asked to apply this machine to the allegedly contemporaneous notes of three interviews with one Paul Dandy, during which he had apparently confessed to the offence of armed robbery. These notes had been taken by the Police-specifically, by members of an elite unit in the local police force: the West Midlands Serious Crime Squad. Of whom, at that time, I had not heard; nor had it ever occurred to me that it might be a good idea to use ESDA on the disputed interview notes that had cropped up from time to time in the course of my practice as a document analyst. [1]

The West Midlands Serious Crime Squad is now famous, and so is ESDA, and their names are inextricably linked. The Squad was disbanded in 1988 pending investigation into a number of allegations of fabrication of interview notes. After that, one by one, a chain of similar alleged miscarriages of justice led to the release of people serving long prison sentences. These included two of the most famous, notorious, crimes of this half of the century: the murder of PC Blakelock in the Broadwater Farm riots and the Birmingham Pub bombings. Before 1987 no-one who was either not a member of or in serious contact with the criminal world took seriously the proposition that the Police were capable of forging confessions. Since the chain of events that I have mentioned, no-one believes anything other than that. At the heart of this extraordinary reversal was ESDA, whose evidence was responsible for most of the successful appeals; and there are those who believe that it was the Dandy case that started this avalanche. Since the Dandy case I have conducted a large number of these analyses of Notes taken by the West Midlands Serious Crime Squad and other police forces, including the Flying Squad, with results that ranged from negligible to dramatic. In this paper I propose to tell part of the story of ESDA and the analysis of Police Interview Notes. I will describe the ESDA machine and its functioning; I will describe the Notes themselves and the rules governing their taking, known as PACE: the Police and Criminal Evidence Act of 1984. I will describe the techniques that have been developed to use ESDA for this purpose, and the problems and pitfalls associated with these techniques. Finally, I will give a detailed account of one case.

ESDA

ESDA is an extraordinary machine. It brings to light evidence that was not only hitherto completely invisible, but whose very existence was unknown. It produces not just identification evidence, but text: evidence with content. It can reveal whole narratives, and these narratives can reveal the crime that has been committed. Paradoxically, since what we are dealing with here is fabricated confessions, what it reveals amounts to a confession of this fabrication: a text beneath the surface that tells the story of how the fabricated text was created. This sometimes has the uncanny effect of taking the investigator back in time, to the occasion, perhaps in some interview room in the West Midlands in 1986, when the document was first written. At its best, the effect of ESDA is almost magical.

What ESDA actually does is very simple. If you write on a piece of paper that is resting on another piece of paper, it is quite likely that as you write the pressure of the writing implement at its point will create indentations on the sheet beneath. If there were carbon paper between the two sheets you could read these indentations as blue or black images; without it, the indentations are at best difficult to read, and usually more or less invisible. ESDA reveals the indentations almost as effectively as carbon paper: if you apply the ESDA process to the apparently blank under-sheet, the machine produces a legible image of these indentations, even if there had been one, two, or even three other pages intervening between the top copy that received the writing in the first place and the page actually under examination. When this happens, the results can be dramatic: suddenly, out of nothing, as it seems, an entire text springs to life.

ESDA was invented in 1978 at the London College of Printing by D.J.Foster and D.J.Morantz, in consultation with the Metropolitan Police Forensic Science Laboratory. [2] It works by an electro-static process, in much the same way as a photocopier. The sheet of paper to be investigated is placed on a bronze slab. This slab is earthed, and sintered: that is, it is perforated by tiny holes. Underneath is a powerful (and noisy) pump, which, when switched on, sucks air through the slab and holds the paper down flat against it. A sheet of transparent polymer film (very similar to the food wrap known in the UK as 'cling-film') is laid over the top of the paper, both to protect it and as an essential but somewhat mysterious part of the process. The resulting sandwich (transparent film, paper, and bronze slab) is then charged with static electricity: a metal wand known as the corona discharge unit is passed over it. Inside the wand is a thin wire with a potential of about 5,000 volts. As a result, static electricity discharges itself to earth through the sandwich.

It seems that the conductivity of the sandwich differs at those places where there are indentations in the paper, and this difference results in an electronic image of the indentations being created on the sheet of film. This image is developed, as it were, by allowing fine carbon granules to spread over it, either by an aerosol process in a cloud chamber, or by inclining the bronze slab at an angle and pouring the carbon granules, mixed with small plastic balls which act as a carrier, across the surface. The granules stick to the electrostatic image of the indentations, and the image becomes visible. It is preserved by putting a sheet of adhesive plastic on top of the layer of carbon, thus creating another sandwich (adhesive plastic, carbon, and transparent film) which can be lifted away from the paper. This sandwich, which has come to be known as an 'ESDA lift', contains the image of the indentations and preserves it very effectively.

An ESDA lift, then, looks like a sheet of thin transparent plastic with an overall grey tone formed by the tiny dots of carbon, in which the image of the indentations is preserved in the form of local coalitions of those dots, which appear as black, or at least darker grey, marks. ESDA does not ignore any top copy writing that there may have been on the sheet under examination, but it treats it in the opposite way from non-inked indentations: top copy writing is developed as white lines against the grey, like a photographic negative, and so is easy to ignore. This aspect in particular of the ESDA process is not fully understood: why is it that the inked indentations of top-copy writing are treated in the opposite way to the non-inked indentations? Theories have been advanced, but none has won universal acceptance.[3]

Nor, in practice, does it matter very much. We do not need to understand how ESDA works to know that it does work, and to have complete faith in the texts that this process reveals. All one has to do to be converted to this belief is to see it in action: it is as immediately convincing as a photocopier. Both processes result in conglomerations of carbon particles that have coalesced to form images. These images are so obviously coherent and significant that even if one didn't have the original source of the images (the copied document, in the case of the photocopier, or the original writing that had produced the indentation, in the case of ESDA) no-one is likely to believe that the preserved text was an artefact produced by the random distribution of carbon particles. It is immediately obvious that the odds against this are almost limitless.

ESDA, then, provides compelling evidence, when it works. Sometimes it doesn't, for whatever reason; perhaps there are no indentations on the sheet under examination, or perhaps the indentations that are there have not reproduced, though some incorrect combination of the distance of the wand from the bronze slab, the surrounding humidity and its effect on the conductivity of the paper, the amount of carbon, or whatever. ESDA is somewhat temperamental, and its silences are deeply frustrating; but when it speaks, what it says carries conviction.

Which is why it has done what it did. In the West Midlands (at least) any criminal lawyer will tell of a long series of complaints against the police for fabricating evidence which have either failed for lack of evidence or, more commonly, not even been brought to Court, for the same reason. An allegation made by a suspected criminal against the forces of law and order has been felt to need particularly convincing evidence to support it. ESDA has been found to provide exactly this. But in order to understand the nature of this evidence, it is necessary to understand the documents on which it has been used: Contemporaneous Notes.

Contemporaneous Notes and PACE

There are strict rules governing the way in which interviews take place between police officers and those they suspect of having committed crimes. These rules are laid down in the Police and Criminal Evidence Act of 1984, known as PACE. These rules were accepted as a Code of Practice early in 1986. What follows is not a description of PACE, because I am not a lawyer, and do not have a lawyer's knowledge of these matters. Rather I will describe the normal process by which Contemporaneous Notes are taken, subject to PACE, from the point of view of the examination of the end product: the Notes themselves.

The actual product of the interview is [4] a set of unbound A4 pages. These are hand written on standard pre-printed forms, according to a method determined partly by the Code of Practice and partly by habit and convenience. This method differs from Police Force to Police Force, but within each the practice is remarkably constant. The formalised procedure that results in the production of these notes has at least three protagonists: the interviewee, the interviewer, and the note-taker. The note-taker takes a standard blank first page and begins by writing, in the allotted spaces, the interview details: the interviewee's name, age, address, the date, the location of the interview, and the names of the officers present. A Caution will then be administered, which the interviewee is invited to sign, in a box on the pre-printed first page. The note-taker will next enter the start time of the interview, first in another box in the middle of the interview details, and then (usually) in the left hand margin of the bottom half of the page where the beginning of the interview proper is to be recorded. The interview then begins, in the form of questions by the interviewer and answers by the interviewee, and each is recorded, as it happens, by the note-taker. There are six lines (in the forms used in the West Midlands) available on the first page for the beginning of the interview; when this space is full, continuation sheets are used, which allow 24 lines of writing. The note-taker will not interrupt the difficult task of contemporaneous note-taking to fill in the heading details and page numbers at the top of each continuation sheet; only the text is written during the interview. The Code of Practice insists that the Notes should be a verbatim contemporaneous record of everything that is said and done in the interview. Interruptions, for tea, for instance, or because of visits by a supervising officer, are timed and noted.

When the interview is over the end time is recorded and entered, sometimes at the end of the interview notes, but always in the box provided on the first page. The notes are then read over to the interviewee, or he or she is invited to read them; during the course of this reading the interviewee is asked to sign every page of the notes and to initial any corrections. The rereading done, the interviewee is (in the West Midlands, at least) asked to write in his or her own handwriting at the end of the interview what is known as the Caption: a short formal statement to the effect that the interview has been properly conducted and the notes are a true record of the interviewee's own words. This is signed, and then the Notes are 'topped and tailed': that is, the note-taker enters the Heading-the full name of the interviewee-at the top of each continuation sheet, and the page number, then then signs the page in the space provided at the bottom; then the interviewing officer also adds his signature at the bottom of the page.

This is what happens from the point of view of the protagonists-when an interview is conducted properly. From the point of view of ESDA analysis, which, it will be remembered, depends entirely on impressions indenting through from a top copy on to sheets placed beneath it, there is one crucial item to be added. The entire success of ESDA in the analysis of interview notes derives from one simple human preference: the fact that people do not like to write on a hard surface. Particularly the hard and not utterly flat surface of the tables found in interview rooms, which have normally seen a good deal of wear and tear. Some form of pad makes a much more satisfactory basis for writing, and, since the pre-printed interview forms come in loose stacks rather than bound pads, the obvious thing to do is to make a squared pile of these sheets and write on them from one to the other, down through a stack. The end result is that while the note-taker was transcribing the interview he or she was also, unknowingly, leaving an invisible duplicate of the written text of each page on the page beneath. From the point of view of this invisible record, then, the following scribal habits are of vital importance. The usual practice is that when the interview begins, the note-taker will take a first page and place it on a stack of continuation sheets. Each page is then filled in using the rest as a pad to support the writing page. When the interview is finished, the normal pattern is for the topping and tailing to be done with all of the sheets in a pile, so that, like the text, the headings, page numbers, and signatures will indent through from one page to the next, though the text writing represents one consecutive writing sequence, and the addition of the details another. Last of all the end time is entered on the first page, often, but not always, while it is resting on a stack of continuation sheets.

When all of this has been done, the Notes are delivered to the Custody Sergeant. This officer plays an important role in the proceedings under PACE. It is his or her task to act as an independent check that the rules of PACE are complied with. Since the Custody Sergeant is based in the Police Station where the interview takes place, and is not a member of the investigating team of detectives, it is assumed that he or she will be independent of any possible attempt to falsify the record. This officer keeps a Custody Sheet, which is a carefully timed account of everything that happens to the suspect in custody, including all interviews and other visits by police officers. It has boxes that should receive all of the details of the suspect, including one box that has the following headings:

I want a solicitor as soon as practicable
I do not want a solicitor at this time

The interviewee is invited to cross out one of these options, and sign it; the entry is timed and dated. When the custody officer receives the Notes, he or she will count the pages and check that they have been headed, signed, and numbered; he or she may also decide to read the Notes. They are then (usually) handed back to the arresting officer to be taken and kept in (for instance) an incident room, together with the other evidence.

If all of the conditions have been complied with, and particularly if the notes have been signed by the interviewee, the resulting pile of sheets becomes evidence: it can be used against the suspect in order to obtain a conviction. But, and here is one of the main causes of contention, if they are not signed they can still be used as evidence, if in a different way: an officer can testify on oath that the interview took place as set down in the notes, and they can be read out and used as evidence. On occasion unsigned contemporaneous notes have been the sole basis for conviction. Using ESDA on contemporaneous notes

I will now give an outline of two of the ways in which we can use ESDA in the analysis of interview notes: specifically, I will describe some of the uses of alignment and misalignment, and the evidence for rewriting. [5]

The first thing to do is to determine whether there are any indentations on the ESDA lift. If not, analysis obviously stops there. It is sometimes tempting to construct a theory on the basis of the non-appearance of indentations in ESDA lifts, but this is a dangerous argument because ESDA may simply have failed to register, for whatever reason, indentations that are in fact present. Many factors combine in indeterminate combinations to produce a good set of ESDA lifts, and it is not uncommon for one analyst to discover more than another, equally competent, expert, simply because of this element of unpredictability.

Usually, however, the ESDA lift will show some record of indentations. The clarity of the realisation of these varies very considerably: sometimes they can be read almost as easily as top-copy writing, but for the most part they fall somewhere in a scale ranging from difficult to impossible. One of the main factors affecting this clarity is the writing implement used. In the West Midlands the Serious Crime Squad made life much easier for analysts of ESDA lifts by preferring in the large majority of cases to use ball-point pens. In the Metropolitan Police, however, there is (I am told) a custom that when an officer is promoted to the C.I.D. he or she will go out and buy a fountain pen. Fountain pens deposit a line of ink without needing much pressure, and the result is often either no detectable indentation at all, or a set of pale grey lines only barely emerging from the surrounding grey fog of an ESDA lift. Attempting to decipher these is not a pleasant activity. All of this is exacerbated by a curious characteristic of ESDA. If it is applied to the indentations of a page of writing, written continuously from top to bottom, it is common that the indentations recorded by ESDA become fainter, often to vanishing point, as they proceed down the page. Why this should be is not known.

Fortunately, however, it is not usually necessary to read the indentations, because they will derive from existing writing: normally from the previous page. So if an ESDA lift has indentations on it, the first thing to do is to find out where they come from. The transparent ESDA lift is superimposed on a photocopy of the previous page. If the indentations derive from this page it is usually very easy to see that this is the case: since the same act of writing produced both top copy and indentations, the match, or register, to borrow a term from printing, between the two will be exact, leaving no doubt as to where the indentations came from. Even very faint or partial writing can be traced to its source in this way, and the need for the labour of reading the indentations is removed.

If the source of the indentations can be determined, then the next matter to concern the examiner is that of alignment. The normal pattern is for the pages to have been squared during the act of writing into a more or less neat pile of sheets: if not, then presumably the pad function of the pile of sheets would be impaired. The result is that when the examiner superimposes the transparent ESDA lift on the source page the indentations will more or less superimpose down the page. But not exactly: during the act of writing the paper will characteristically have shifted slightly as the hand moves downwards in the act of writing, disturbing the relationship between the top page and the one underneath. As a result, the entire set of indentations will not be in one register, but in several different registers of up to as much as ten lines or so each. The heading details (the interviewee's name and page number), since they are usually added after the end of the interview in a separate act of writing, will also not be in register with the text, and the three signatures at the bottom (interviewer, interviewee, and note-taker) will for the same reason not be in register with the text or indeed with each other, since there will inevitably have been a shift in the top page against the pad between the different signings.

This is the normal pattern of alignment. Disturbances in that normal pattern are what the examiner is looking for. For instance, in one Interview I have examined three consecutive lines read:

R I was standing by the phone box keeping watch
Q What did you all do then
R We ran back to the car

These lines indented on to the next page, and the first part of the first line and all of the second line and third lines were in one exact register with the indentations they made. However, crucially, the words 'keeping watch' were out of this alignment. The result can be shown in a stylised way in this diagram:

The bold type shows the exact superimposition of original and indentation; the contrast with the non-register of 'keeping watch' is even more striking in the original. There are two possible ways of explaining it: either the words were written in the correct sequence and the paper shifted twice, the second shift bringing it exactly back to where it was relative to the sheet underneath before the first shift took place; or else the words 'keeping watch' are a later insertion. The first possibility is so unlikely that this kind of evidence is taken (and in this instance, was taken) as a strong indication that the misaligned words were added afterwards, and out of sequence. As the Lord Chief Justice put it in the Court of Appeal, when this evidence was presented to him, 'it doesn't inspire confidence, does it?'

Note, however, that in this case, as usually in the interpretation of ESDA material, there are two components. One is to do with the physical evidence: the disposition of marks on a sheet of paper and an ESDA lift. The other is to do with content. The physical evidence in this instance is strong that the two words 'keeping watch' were added at some point after the aligned text was written: either immediately after, or at some undetermined point after that. But one cannot ignore the content: if the added text is trivial, then its significance is also trivial: an innocent afterthought, perhaps, added a few lines later. PACE requires that any additions or corrections in the interview should be initialed by the interviewee, but it would look absurd to initial an invisible correction. I have seen a number of examples of minor additions and corrections revealed by ESDA which have not been initialed, and no-one would consider making an issue of this. However, in this case the addition is quite evidently not trivial: 'I was standing by the phone box' is not incriminating, but with the addition of 'keeping watch' it becomes an admission to a crime. Normally the task of the document examiner is to make statements about the physical characteristics of the document; its content is the province of the lawyers who conduct the case, and the Court who will determine its outcome. Document examiners don't spend much time reading the documents they examine. But the special ability of ESDA is to reveal new textual material, evidence with content, whose significance can only be determined by reading it, and this has, I suggest, the inevitable effect of bringing the document examiner into the difficult area of textual interpretation.

Another use of the study of alignments is in the examination of start and finish times. Normal practice is to record the start time at the appropriate place in the interview details on the first page, record the interview, and then enter the finishing time, again on the first page. This is usually shown in the ESDA lifts, in that the finishing time will be out of alignment with the text that surrounds it, showing that it is a later insertion. If however the finishing time is in exact register with the start time and the other heading details then this tiny fact is, potentially, of momentous consequence: it suggests that when the interview details were filled in the note-taker knew when the interview was going to end. Which is impossible, if the subsequent pages are a contemporaneous note of an interview that had not yet taken place. Therefore the evidence suggests that the entire set of notes is a fabrication.

The problem here is that this is a major conclusion from a tiny detail; a detail, moreover, that it is sometimes hard to be certain about. Misalignment is obvious: alignment, not so. ESDA lifts are fuzzy, for one thing; another problem is that one must be aware of the edge of the paper: it is a natural action to square a stack of sheets by dropping the bottom and then the sides against the flat surface of the table. If one alignment is square relative to these two edges, then it would be possible (though, I suspect, difficult) to recreate the same alignment by this squaring process at some later point.

If the indentations do not align with anything on the previous page then the usual circumstance is that they are in a different handwriting, and derive from some other interview. One common occurrence is that the note-taker has come to the end of a pile of sheets and started on a fresh pile, the top of which had previously been underneath when a different interview was transcribed. In these cases it is (unfortunately for the eyesight of the analyst) important to transcribe as much of the text thus revealed as possible. The other interview can often be traced, and its date and time determined. If it can be dated significantly after the date of the interview under examination this would obviously give cause for suspicion that the notes of the latter were not written on the date when the interview took place, which would be good grounds for suspecting fabrication.

If, however, the indented writing is in the same hand as the rest of the interview, but does not derive from the previous page, then it must be traced to its origin. Sometimes it may originate from top copy writing in a different interview that is part of a set of interviews of the same suspect. Tracing this origin through perhaps dozens or even hundreds of pages is not easy by hand, but there is a simple solution to this problem: the interviews can be transcribed into a word processor, and an ordinary computer search routine will either find the source immediately, or show with some definiteness that the text is not to be found in that set of notes. The computer can effortlessly trace all of the occurrences of even a few letters in part of a word, and, if care is taken to circumvent the possibility of mistranscription, even fragments can be sourced in this way. Once the source is found, superimposition will (as usual) often reveal the meaning and origin of indentations otherwise too faint to read.

If the indentations do derive from some other interview in the same sequence, then one has an anomaly that can sometimes suggest tampering, if other evidence supports this possibility. Essentially there are two possibilities. It is not uncommon for officers to have the notes of an earlier interview present when conducting an interview of the same suspect, though it would be unusual, I imagine, for these notes to find their way into the pile of blank sheets that is normally used as the writing support and thus receive indentations. On the other hand, to concoct an interview it is essential that it should be consistent with other interviews, and therefore very desirable to have other Notes available for cross-reference. Of course this does not in itself prove anything; however, I can say that I have not seen a set of notes in which this phenomenon occurred without there having been other evidence of tampering to go with it.

If the indented writing cannot be made to superimpose on any writing that belongs to any related interviews, but is in the same hand as the interview under examination, then the possibility of rewriting arises. The evidence for rewriting must be a matter of interpretation of content. The physical evidence is that indentations occur which are in the same hand as that of the note-taker of the notes under examination, but cannot be traced to any extant writings that one has. This in itself, however, is not necessarily significant. What is important is that the indented writing should have a clearly similar content to some piece of top copy writing, but not derive from the same act of writing that produced that top copy writing. If so, then we can suggest revision.

Though unacknowledged rewriting is necessarily a contravention of the PACE rules, it need not be a serious one: occasionally, for instance, two or three lines of indented writing are found at the top of a page that are almost exactly the same, with a small variation. In one set of notes that I examined the following indentations were found:

throw any bombs or prepare them?
I had three bottles of milk I gave them
Stix & he poured the milk out

The indented text ended abruptly at this point. The writing was in the same hand as that of the note-taker of the notes in question, and was in register with nothing on the previous or any other page; but on the page where the indentations were found the first three lines read as follows:

throw any bombs or prepare them?
I had three bottles of milk, I gave them
to Stix's mate & he poured the milk out

How does one interpret this material? A reasonable, indeed obvious, assumption is that the interviewee said the indented text, and had it recorded so that it indented into the next blank sheet. He then changed his mind: not Stix, but Stix's mate was the one who poured the milk out. The recording officer, perhaps in recognition of the importance of this change, instead of recording it and having it initialed, since only three lines on an otherwise clean page had been written, simply threw away the early version and began again on the next blank sheet in the pile-unaware that this sheet had received the indentations of the first attempt. Thus this small contravention of PACE was preserved, invisibly, until ESDA brought it to light.

This is a simple example of the analytic method that one uses in investigating ESDA evidence, but it shows very clearly the process, and the fascination, of the method. There are again three components: the physical evidence, the interpretation of that evidence, and the use of that interpretation for a suggested reconstruction. If the reconstruction is as plausible as this one, then the fascination is evident: an intellectual puzzle is solved, and a hidden piece of the past suddenly, unexpectedly, crystallises into existence.

Not all rewritings are as innocuous as this one. This, finally, is what has made ESDA famous: when it brings to light what can be taken as an early and discarded version, just as in the above example-but with a crucial difference: the current version contains incriminating material that is lacking in the newly discovered early version. In the case of Paul Dandy the key evidence against him consisted of two admissions allegedly made during one interview. Since Dandy vehemently protested his innocence, the admissions were contested. The second of the two admissions occurs in the following passage: it is printed in bold type.

[WER] WILL YOU SIGN AN AUTHORITY
FOR US TO LOOK AT YOUR BANK ACCOUNT?
NO'
I TAKE IT FROM YOUR EARLIER REPLY
THAT YOU ARE ADMITTING BEEN INVOLVED
IN THE ROBBERY AT THE MEB?
YOU'SE ARE GOOD, THURSDAY, FRIDAY,
SATURDAY, SUNDAY AND YOU'VE CAUGHT
ME NOW YOU'VE GOT TO PROVE IT
DO YOU WANT TO READ OVER THE
[NOTES SIGN] AND CAPTION THEM
ILL INITIAL THE MISTAKES BUT
I WON'T SIGN THEM

On the last page of an earlier interview with Dandy the following indentations were found (square brackets indicate conjectural readings):

WILL YOU SIGN AN AUTHORITY
FOR US TO LOOK AT YOUR BANK [ACCOUNT]
NO
DO YOU WANT TO READ OVER [THE]
[NOTES SIGN] AND CAPTION THEM
I'LL INITIAL THE MISTAKES BUT
[I] WONT SIGN THEM

The similarity of content, together with the physical evidence, shows beyond reasonable doubt that another version of an extant page existed at some point, and moreover that this version lacked the incriminating pair of sentences printed in bold type. A reasonable reconstruction would suggest that the indented version was earlier, and discarded in favour of a later version that was offered in evidence, which had the incriminating sentences added to it. Whether this was done without, or, as the officers involved were later to allege, with the consent of Paul Dandy, cannot be determined from this evidence alone. But the possibility of its having been done without his knowledge was strong enough to persuade the judge in his trial to dismiss the case against him immediately.

What I have done so far in this account is to outline some of the basic strategies that one uses in examining ESDA lifts of contested contemporaneous notes. But that is not where the fascination of this work lies. Each case is a unique puzzle: an enigma, that can, if one is lucky enough, be forced or cajoled into revealing its truth. To illustrate this I will now give an account of an actual case in detail.

The Lewis case

George Glen Lewis heard of my work with ESDA and interview notes in February 1991. He phoned me from prison and asked me to help him; he was serving two concurrent sentences amounting to ten years for three robberies, all on the basis of two uncorroborated confessions, the first of which was obtained by members of the West Midlands Serious Crime Squad. This, as I understand it, is his story. Mr Lewis lived in Wolverhampton, a small town in the Black Country near Birmingham. In January 1987 his car, he said, was stolen, and he reported this theft to the local Police. They told him that the car had been found, and he went down to the Police Station to pick it up. There he discovered that his car had been used in a robbery, and that he was the prime suspect. He was arrested, protesting vigorously, and taken to another Wolverhampton Police Station, Red Lion Street, for questioning. He refused point blank to answer any questions without his solicitor being present, and in general contested the whole business with some force: there seems to have been a fight at one part of the proceedings.

I began to shout and swear at the Police officers and I was told to go through into the back charge office. I refused, and eventually two officers came and grabbed me and tried to push me through into the Office. I resisted, and eventually other officers came over and I was overpowered. I was handcuffed on the floor and then taken through into the back office and put in a cell.

When his solicitor arrived he made a statement denying any involvement with the burglary, and was released on bail without charge, with an undertaking to return at a later date.

The following Wednesday two members of the West Midlands Serious Crime Squad found him in a friend's flat and arrested him again. In the car he was, he says, 'verbally and physically abused'.

In the Police Station one of the arresting officers took several blank statement sheets from a drawer and told me to sign them. This I refused to do. He punched me on the back of the head. I still would not sign the blank sheets. Another officer whom I do not know and cannot name pulled out a syringe needle and a phial with a clear liquid in it. He said 'This'll make the black bastard talk'. ... I honestly believed he was going to drug me with the syringe. I am absolutely scared of needles and the sight of the syringe worried me. I then said that I would sign whatever statement paper they wanted me to. [They] gave me a pen and stood behind me looking over my shoulder saying 'sign there', and pointing at the parts where he wanted me to sign ... Bearing in mind that no-one knew where I was, I was under no illusions that these officers would seriously harm me.

The Police officers then, he claims, in front of him, wrote out two confessions to three separate robberies.

I was crying and upset and kept on saying 'Why are you doing this to me?' They said 'Because you fucked them about last week' ... [they] kept going on about how I had fucked them about last week and that I had now been well and truly fucked.

The interviews as recorded in the Notes of course present quite a different picture. In them Mr Lewis decides from the beginning to confess everything. Given the fact that a week earlier he had adamantly refused to say anything without a lawyer being present, the officers ask him why:

DS Firstly let's talk about the burglary that happened last week at Tettenhall.
A The Police have already interviewed me about it. I did it.
DS Well tell me why are you admitting this to us when you denied it previously.
A Look you're in the Serious Crime Squad. I don't want to fall out with you lot but I will tell you about everything. DS What do you mean everything
A Look I'll tell you what you want to know.

When the Notes were complete Mr Lewis was allowed (in his account) to summon his lawyer. This is the lawyer's story:

I requested to see my client but I was kept waiting for a period of approximately 10 minutes. When I was admitted through to the charge area the officers involved in the case were just scuttling out of the back entrance ... [Mr Lewis] was tearful and his first words to me were Mr Dawson they've fitted me up with burglaries and robberies which I didn't commit. I immediately asked him why he had not sent for me immediately when he was arrested. I assumed from his experience the previous week at Red Lion Street he would know to do this. He said that he had wanted to send for me and that when he was deleting the line "I do not want a solicitor; his hand had been knocked aside so that he was unable to complete the deletion. I immediately left the room and went to examine the prisoner in custody sheet and saw that what he had told me was true.

As indeed it was: the custody record has a horizontal line drawn through the first three words of the printed sentence 'I do not want a solicitor at this time', as if to begin to erase them; the line then angles sharply downwards for about 1 cm. David Baxendale, the Home Office expert who was asked to examine this material, commented that this evidence was 'in keeping with the writer's hand being knocked away when the ... line was being drawn.' This evidence (though not Dr Baxendale's interpretation of it) was used in Mr Lewis's trial, and in a complaint that was made to the Police Complaints Authority. The complaint was unsuccessful, as was Mr Lewis's defence at the trial: he was sentenced to ten years imprisonment, almost entirely on the basis of these allegedly false confessions. He was subsequently refused leave to appeal.

Two years later the West Midlands Serious Crime Squad was disbanded, in disgrace, and a massive investigation was conducted into all of its disputed cases, among them that of Mr Lewis. David Baxendale was asked to examine the ESDA evidence of the two interviews. His conclusion was that the ESDA evidence showed nothing of any significance. In the first interview the text had not indented through from one page to the next: there were indentations of signatures and page numbers, as part of the 'topping and tailing' process, and the last page bore some indentations from another interview, but nothing else. It is reasonable to assume that each page had been written out while resting on the table top, or some other pad, and not on a pile of blank sheets: unusual, but not of any interest. As he said: 'The evidence as to whether or not these notes were compiled in an irregular manner is essentially inconclusive'.

The second interview had more material, including some anomalies. The text of each page indented on to the next page, in the normal fashion, but the last page had no indentations from the page before. This is odd, but not necessarily of any significance at all. The heading details and police officers' signatures, where they could be deciphered, appeared also to indent in a regular fashion, from one page to the next. The other anomaly, however, was in the pattern of indentation of the Lewis signatures. He found the following: the Lewis signature on page 2 indented on to page 3, as one would expect, but that on page 6 derived from the previous page, page 5; similarly the page 6 signature indented on to page 5. He found no Lewis signature indented on to page 6, but on page 7 he found a Lewis signature indented that derived from an original on page 3. This is odd, but not itself evidence that casts doubt on the authenticity of the Notes: it could simply be an inexplicable anomaly, the product of some unique set of circumstances: the notes were shuffled, for whatever reason, when Lewis signed them. Accordingly, he concluded that there was no evidence to suggest that the notes had been tampered with.

That was where the matter stood when I received the notes for examination in 1991. [6] I took fresh ESDA lifts, and examined the ones that David Baxendale had used: essentially, I agreed with his report. Was there, then, no more to be said? It was hard to leave it at that. Not just the vivid circumstantial detail of Mr Lewis's account, or the obvious misgivings of his lawyer, or the evidence that suggested that he had been prevented at one point in the proceedings from seeing a solicitor; there was a feeling that somewhere in the ESDA prints was a solution, if only one could find it. My two colleagues, Kelsey Thornton and Claire Marshall, and I spent many hours wrestling with this enigma. Eventually the following hypothesis emerged.

In interview two the text of the questions and answers indents from one page to the next in a regular fashion, with the exception of the last page, which has no indentations from the text of the preceding page.

The pattern of the signatures was, however, different. Prolonged examination of the vague patterns of grey dots that hover on the edge of legibility revealed the following. It was possible to add to the indentations found by David Baxendale to make a more complete picture of the pattern of indentations of the signatures. In my report I wrote:

The signature from page 2.2 [7] indents on to 2.3, as would be normal, but that on 2.3 indents on to 2.7. The Lewis signature on 2.6 indents on to 2.5, and that on 2.5 on to 2.4. The signature indented on 2.2 is not clear enough to determine its provenance exactly, but it is not inconsistent with the possibility that it was produced by the signature on 2.4. This suggests that when the pages were signed by Mr Lewis they were in the following order:

2.6 : 2.5 : 2.4 : 2.2 : 2.3 : 2.7

This new order (which was agreed by David Baxendale in a comment on my report) is interesting, because it is beginning to look like a pattern:

The pattern is that the new order is almost a reverse of the order of the text pages. Now, Mr Lewis says that he was induced into signing a number of blank sheets. Supposing, as he signed each one, he placed it face up on top of the previous one, including 2.7; but he also, as well as signing it, wrote a caption on 2.7, thus producing a pile of sheets in reverse order. Separately, on a page from a pile of blank first pages, he signed 2.1, whose signature is not found indented anywhere. This gives us the following sequence:

2.7 : 2.6 : 2.5 : 2.4 : 2.3 : 2.2

with 2.1 kept separate because it is on a pre-printed first page form. In the course of writing the notes out, the officers would first fill in page 1, and then take the pile page by page in the above order. The top page, 2.7, since it had a caption on it and would therefore have to be the last page, would be put aside, and the rest filled in the order presented in the pile, except that 2.2 and 2.3 somehow got transposed. Finally 2.7 was filled in, and the Notes were then topped and tailed as usual.

There is no doubt, I think, that this is an interesting hypothesis. It is only that, of course. It is not an opinion as to what took place in Wednesfield Police Station between 7.55 and 8.35 pm on January 21st 1987. It is a possible reconstruction: what the Police call a 'scenario'. It has strengths and weaknesses. It strength is that it accounts for all of the evidence at hand, including both the ESDA evidence and Mr Lewis's version of events. This latter point is a considerable strength, since his story can be said to have predicted the evidence of ESDA. It does not speculate, in the strictest sense of the word, since it does not make up extra items of the story to account for inconvenient anomalies in the data-except in one instance, which is its greatest flaw: it is forced to rely on chance to explain the transposition of 2.3 and 2.2. Its weakness is simply that it does not compel assent: the natural reaction would be perhaps to applaud its ingenuity (if one were feeling generous) and to say, yes, it could have happened that way, but it need not have done.

In my report I wrote:

This reconstruction is of course highly conjectural. It is however coherent, and it explains the evidence of the ESDA indentations, which are otherwise hard to account for.

David Baxendale was not impressed:

The pattern of the impressions of the "Lewis" signatures ... could also have an innocent explanation. For example, Lewis could have jumbled up the sheets while he was reading them, and then signed them without putting them back in order.

This is indeed true, but as an alternative hypothesis it leaves something to be desired. It does not address itself to the specific phenomena under examination, namely the actual and particular order to the signings as revealed by ESDA. It says that that order could have happened by chance. This is true, but it is trivially true: anything can happen by chance. The number of possible chance shufflings of seven pages are seven factorial: 80640 possible combinations. So the odds against these seven pages falling by chance into that particular order (or any particular order) are 80640 : 1 against. [6] At least. It could be argued that we are not talking about one sequence of seven pages (the signing order) but about two linked sequences: the text order and the signing order. Both these sequences are explained by the Davis hypothesis. To explain them by chance (the Baxendale hypothesis) would be to maintain that chance determined these two sequences out of a possible 7 factorial squared other combinations that chance could also, with complete impartiality, have provided. In this view, the odds against the Baxendale theory are 6,502,809,600 : 1 against. I would maintain, therefore, that the Davis theory, though weak in the ways described, is the best there is, in that no better has been proposed.

The question is, once it has arisen, what does one do with it? To propose hypotheses for refutation is a perfectly valid scientific procedure; indeed, it is generally maintained that this is the defining procedure of science. But in this case its merits and demerits are ultimately a matter not for science, but for law, which operates different criteria when determining the value of a theory. It is the task of the Defence to demonstrate to the satisfaction of a Court that there is reasonable doubt, and no more than that, about the guilt of an individual. 'Reasonable doubt' seems to be a rather elastic concept, perhaps deliberately or at least necessarily so: it seems to be, in the last analysis, determined for individual cases by what happens to those cases in particular Courtrooms on particular days. I would suggest that, given that elasticity, it is the task of a scientific expert working for the Defence to point out to the lawyers employing him or her any possible interpretation of the phenomena that may possibly be of assistance in establishing reasonable doubt. It is then for the barrister to judge as to the utility of this material for his or her task, and the Court to judge as to its final value.

I sent my report to the Defence on 21st May 1991. They sent it to the Home Secretary, who gave Mr Lewis leave to appeal. On March 24th 1992 Lord Lane in the Court of Appeal ordered a retrial. This retrial took place in the week beginning July 20th 1992; on July 22nd the Judge ruled that the confession evidence was inadmissible, the prosecution offered no further evidence, and Mr Lewis was found not guilty. He had served seven years of his ten year sentence.

Tom Davis 11/9/92