Police research series - The Alpha papers

Research affecting the reliability of Verbal Evidence in UK Courts.

Don't trust the evidence of your ears.


1. Introduction

This series of investigations was aimed at exploring the potential problems of automating Command and Control operations in the Police and elsewhere. However this element of the research touched a much more sensitive nerve, namely the veracity of verbal evidence in our justice system.

The overall inquiry was designed to investigate the abilities and requirements of Police Command and Control Operators. It is part of a series of over 20 investigations into this subject, prompted by the imminent introduction of computer systems into this area of police work.

An important result was discovered which strikes at the very heart of evidence in the British judicial system. This is a matter which whilst generally important was peripheral to the main thrust of this research effort. It is therefore reported here as an ancillary report to the main work as it has ramifications far beyond the design and operation of Command and Control Systems


2. Summary

My research, well supported by psychological precedents proves that no reliance can be placed upon recollections of actual words spoken by witnesses at the time of an incident, even when honest and diligent professional persons contemporaneously record these matters.

The problem is that the human brain is not designed as a tape recorder and it is easy to show that even professional witnesses cannot provide an unaltered account of the real words used (see Bibliography). This is an important finding as frequently criminal cases can turn on the 'actual words ' spoken in a particular situation. These findings should provoke a review of the rules relating to verbatim evidence in UK courts.


3. Introduction

This research was undertaken at the time when computers were being designed for Police control room use. A control room is invariably the first point of contact for the police service with a violent or unlawful incident.
Normally the sequence of events are that such an incident is discovered by a member of the public, very few are discovered by the police themselves.

That member of the public or another witness then contacts the police service. Generally in this day and age that contact is by telephone, the majority of which go into police emergency control rooms.

The personnel in these control rooms are responsible for obtaining accurate details of any incident from that witness and then deciding on appropriate action. If that action involves the dispatching of resources they are also responsible for briefing the attending officers, assuring adequate resources and monitoring the progress of the situation etc.

As noted in some other research monographs in this series it must be emphasised that these operators regularly work under high stress conditions involving often confused and in some cases deliberately falsified information. They are dealing with witnesses under emotional tension frequently in less than optimal conditions for the easy comprehension of the message being passed to them.

In most cases the police operators write down scrap notes of the conversation with the emergency call witness during the course of the telephone call. Later, usually within 15 minutes, they transpose those 'notes' onto an official emergency incident log. That log then becomes the first official record of the reporting of the incident, it is customarily thereafter the basis of any subsequent police actions.

As noted during the observational phase (see observational research monograph) of this investigation the operators wrote down the majority of the emergency message logs as though they were verbatim reports of the actual words said in the emergency call to the police. In other words the written report purports to be an exact transcription of the actual words used by the witnesses or sometimes offenders during the emergency telephone calls.

This is crucially important in the wider context of criminal evidence but not necessarily too important in the situation of my investigation into the capture of meaning in an emergency situation. In the latter case the police operators are simply interested in getting enough accurate information to decide upon the selection, allocation and briefing of the initial emergency response personnel from their police agency or from firefighters, ambulance, coastguards, lifeboats, military or local authorities etc. The exact words do not matter too much as long as the professional police emergency operator captures the essential meaning. (Whether that is the case is further investigated in another inquiry monograph in this series).

In the former criminal evidence case the situation is entirely different, here the actual words used, rather than the interpreted meaning are often of vital importance. Largely because the lawyers like to place their own meaning on the words used, normally to press the probability of the interpretation of the events that they are currently peddling.

Not infrequently QCs (lawyers, attorneys) will argue at length over the nuances of a single reported word. A person's liberty or even life could hang on whether a certain word was used or whether it was another similar word that could be capable of an alternative meaning. Frequently 'Incident Logs' and other early incident recording forms end up in the courts as evidential documents.

In view of the fact that the majority of the emergency logs were written down as though they were such 'evidence' I was interested to know how accurate that reporting might be. Particularly as they often ended up in the higher courts of the land, being minutely dissected by prosecution and defence lawyers.

I once had a case where my pocket book was in the hands of the judge and the defence lawyers for over 90% of the time that I gave evidence. (Later I wrote all my pocket books in an esoteric shorthand to prevent being hoist by my own petard, but that is another story!)


How secure a foundation are these 'accurate' notes made at the time for evidential use in courts worldwide?


Most verbatim evidence recording is not capable of investigation. It occurs in the street or at a scene of crime when a police constable or a detective makes notes in his 'official' notebook. At that point there is no way of checking the accuracy of the words noted down versus the words actually uttered in the situation. "It's a fair cop guv" has become infamous in the annals of police reporting; leading to the hypothesis that all criminals in London (UK) first undertook a course at RADA.


So how accurate is police reporting of the 'actual words' said by a witness or suspect.


4. Investigation

I took a random sample of 190 emergency call messages for a much larger pool of about 2000 available messages. Each of these calls had been audio recorded without the knowledge of the operators. Each audio call was transcribed and the transcription checked by three independent police officers each of whom listened to the recording and checked the written transcription. This was to ensure that the transcription was as accurate as possible.

Every related 'Message Log'; that is the written record of the emergency call; was secured and transcribed into a computer system. Each of these written transcriptions was checked by another (different) set of judges to ensure transcription accuracy.

Each message log was then placed alongside the related audio transcription and two senior expert police judges working in concert checked every call transcription against every message log. Each of these logs purported to record the exact information in the emergency call. (Although it is not really relevant in the terms of research evidence, three researchers also checked the logs and came to the same conclusion as the expert judges)

The judges rated the calls firstly on whether they were in a verbatim (actual words used by the witness) or a reporting style.

Then for the verbatim style messages (180 of them) they rated the accuracy of the reported surface structure, that is to say the actual words used.


For the verbatim messages they rated whether :-

  • · It was an exact replication in total of the emergency call
  • · If any clause exactly in the call matched any clause in the log purportedly recording the call
  • · If any clause in the call nearly matched a clause in the message log. (Nearly being defined as no more than five words being changed or transposed)

** A clause is a part of a sentence that contains a significant assertion. In other words it is a fragment of a sentence.


The results of their findings are encapsulated in the following table.


Figure 1 Surface structure comparison of '999' calls transcripts and message logs

Non Verbatim Style
Verbatim Style
Number of logs which contained a single clause which had:
An Exact Translation
Some surface similarity
NO surface similarity


It can be seen that of the 190 messages 180 or nearly 95% were in a verbatim style, purporting to be the actual words used by the witness.Of these NOT ONE of the logs exactly replicated the conversation and only TWO replicated even a small part of such conversations.The number of message logs that contained a clause that exactly matched a clause in the call was only two (in this case this amounted to only two logs as there was only one identical clause in each of these 'more accurate' logs)

The number of message logs, which contained just a single clause, which was judged to be similar to the call according the criteria above, was 46 or only 26%.

The majority (73%) of the message logs which purported to be an exact reproduction of the words spoken by the witnesses showed no appreciable relationship to the actual words used in the emergency call conversation.



It would seem that the police operator transposes the information that eventually reaches the message log from the information that he/she received in the call, into some linguistic style of his/her own.

In terms of the main focus of this research this is not necessarily a problem, provided that the essential meaning extracted from each caller is accurately deposited into the computer database (or the paper message log in a manual system) and then subsequently used to effectively brief and deploy police and other emergency resources. Whether this is the case is explored in another of these research monographs, unfortunately with not too happy a conclusions.

The reasons for changes in both surface structure, 'word' recollection and as discussed elsewhere in 'recalled meaning' do not lie in operator mendacity, but in the nature of human cognitive processes and memory. The human brain is designed to make sense of the inputs that it receives and this means that what it experiences is not pure uncut sensory data, but mediated knowledge, effected by existing expectations and world models.

This goes further than saying that we receive sensory input and the interpret it, what I am saying is that the sensory input that we receive is already effected by our current and existing expectations, we see or hear what we expect to see and hear. A more in depth discussion of this somewhat disturbing thesis is undertaken in another of these monographs.

In this report I want to return to the more practical aspects of our findings.

Traditionally a great deal of weight has been placed upon records of this nature, which are (supposedly at least) made at the time of an incident or its report. When written in a verbatim style they are generally regarded by all parties as being an accurate record of the words actually used at the time, especially when the veracity of the witness is not being called into question.

In most discourses it does not matter very much if the surface structure of an utterance is changed drastically as long as the essential meaning of the dialogue is retained. However in many court cases different interpretations are often placed upon minutely dissimilar surface structures of words 'used at the time'. Making the accurate recording of the exact words important in such situations.

These findings confirm what has been well understood by psychologists for a along time, Human beings do not work well as tape recorders. However I doubt if even psychologists realised how bad human beings were at this task and how serious a spanner that throws into British evidential procedures.

The irrevocable conclusion must be that virtually no reliance can be placed upon a witness's recollection of the 'actual words' spoken at any remembered instance even if they made notes at the time. The psychological theories underpinning these findings may be found in the bibliography at (Evidence Bibliography) for those interested enough to research these findings further.

It must be emphasised that I am not talking about the position where witnesses are deliberately lying. What this research concludes, well supported by previous studies, is that even when an honest, sober and competent witness tries to recall the exact words that are used in a conversation they are generally in error. This is due to the way that the brain works. It can be defined as an 'effort after meaning' rather than an audio or pictorial recording of events.

Courts must not therefore accept, as a matter of simple scientific fact, that human recollections of actual words used in any incidents are in any way accurate in so far as the surface structure is concerned. To do so would be to potentially promulgate serious and unintended miscarriages of justice.

(c) John Hulbert 1976