Search

We have written this guide in response to concerns lawyers have voiced about their knowledge of, and ability to, explore the underlying quality of psychological opinion evidence. This guide addresses issues that attentive psychologists and mainstream scientists regard as important for understanding the probative value of psychological opinions .

Examination of cases citing IMM v The Queen

The broad aim of this project is to turn fingerprint novices into experts more quickly by developing more rigorous training practices. To do this, we first have to understand what makes an expert so special. Expert fingerprint examiners are highly accurate in their judgements and consistently outperform novices across a variety of tasks. They can distinguish between matching and non-matching with a high degree of accuracy, even under time constraints and with reduced visual information (Thompson, Tangen & McCarthy, 2014). The current project aims to capture to essence of expert performance in forensic fingerprint examination using think-aloud protocol analyses. To do this, we will analyse and compare the concurrent and retrospective verbalisations of experts novices during their performance on a domain representative task. Using their verbalised thoughts, we will then be able to generate hypotheses for the cognitive processes and mechanisms mediating the superior performance. These hypotheses will be tested in designed experiments during subsequent phases of the project to determine whether the mechanisms are indeed responsible for the superiority of expert performance. This research as a whole will add to the advancement of fingerprint expertise, as careful analysis of task characteristics, performance, and underlying knowledge and cognitive processing is at the heart of improving current practices.

This component includes our complete data analytic pipeline: from raw individual .txt files to final R Markdown plots and analyses. The "analysis folder" contains data files (.txt) for 12 simulated novice participants and 12 simulated expert participants with pseudorandomly generated response data—these are used to test out the experimental analysis script, and model the a correct response rate of 50% (e.g., see the simdata.jpeg plots as illustration). Also included is a LiveCode data extraction tool for aggregating raw data from the individual .txt files, an R Markdown folder with source code, and a fully documented html file documenting all code, notes, plots and results of statistical analyses. A ‘final analysis’ folder will be uploaded here once data collection is completed and the human data have been run through this analytic pipeline.

This component contains the LiveCode source file (version 9.0.2) for presenting the sequences and materials to participants and collecting their responses. LiveCode’s Community Edition, is a free open source object-oriented, cross-platform, natural language rapid development environment that can be used for programming local (LiveCode) and web-based (HTML5) experiments. LiveCode is based on the Transcript programming language (a high-level xTalk scripting language like HyperCard’s HyperTalk), and is a great stepping stone to learning Javascript and Python. It is fit for developing experiments that require fine-grained control, complex sequencing, and heavy back-end processing. You can download LiveCode for PC, Linux or Mac at: https://downloads.livecode.com/livecode. You can run this experiment on your own computer by installing the appropriate version of LiveCode Community, and downloading the zip file. Note, however, that the “images” file is missing, which you’ll need for the experiment to run properly. Please contact Jason Tangen (uqjtange@uq.edu.au). If you have permission to use these images, he’ll send you a link to the images we used in this project. Once you get the “images” folder, just put it inside the main experiment folder, which can be stored anywhere on your machine. But the file structure within this folder needs to be preserved to run the experiment, as the script is reading in the participant sequences, images, and instructions on the fly from the same location as the source file.

This component contains the raw individual .txt files outputted and time stamped on completing the experiment.

This component contains the LiveCode source file (version 9.0.2) for presenting the sequences and materials to participants and collecting their responses. LiveCode’s Community Edition, is a free open source object-oriented, cross-platform, natural language rapid development environment that can be used for programming local (LiveCode) and web-based (HTML5) experiments. LiveCode is based on the Transcript programming language (a high-level xTalk scripting language like HyperCard’s HyperTalk), and is a great stepping stone to learning Javascript and Python. It is fit for developing experiments that require fine-grained control, complex sequencing, and heavy back-end processing. You can download LiveCode for PC, Linux or Mac at: https://downloads.livecode.com/livecode. You can run this experiment on your own computer by installing the appropriate version of LiveCode Community, and downloading the zip file. Note, however, that the “images” file is missing, which you’ll need for the experiment to run properly. As we describe in the Materials component, the fingerprint images we used in this project were sourced from the NIST Special Database 300 (https://www.nist.gov/itl/iad/image-group/nist-special-database-300), which consists of 8871 rolled fingerprints, which were donated to NIST by the Federal Bureau of Investigation. Unfortunately, NIST has not made this special database openly available, and we do not have permission to distribute it further, publish, copy, or disseminate it in any way or form whatsoever, whether for profit or not. If you’d like to access the images we used in this project, contact fingerprint_data@nist.gov to get permission from NIST to use the database (it’s freely available for biometrics related research, development, and education), forward Jason Tangen (uqjtange@uq.edu.au) the automated access email, and he’ll send you a link to the images we used in this project. Once you get the “images” folder, just put it inside the main experiment folder, which can be stored anywhere on your machine. But the file structure within this folder needs to be preserved to run the experiment, as the script is reading in the participant sequences, images, and instructions on the fly from the same location as the source file.

This component contains the raw individual .txt files outputted and time stamped after completing the experiment.

In this experiment, we will test to see whether expert fingerprint examiners have better visual search abilities compared to novices with a class of stimuli they are familiar with. We will also test whether their superior performance disappears when the structural regularities of the stimuli are removed, or if asked to spot information that is non-diagnostic. We have conceptualised a fingerprint-like visual search task much like a Where's Wally puzzle that assesses how well participants can find points of correspondence (the find-the-fragment task). Participants will be asked to find a small fragment of print information (presented on the left) within a larger fingerprint image (presented on the right) as quickly as they can.

The broad aim of our current research program is to turn fingerprint novices into experts more quickly by developing more rigorous training practices, but we first need to identify the leading practitioners in the field so we have a clear picture of what makes them exceptional. In this project, we have developed a fingerprint matching task that’s sufficiently difficult to help separate examiners with different (extremely high) levels of competency. We have created 48 pairs of fingerprints (24 from the same print and 24 from different prints), from actual case files, which have been selected to be extremely difficult to distinguish. These prints were developed in collaboration with our partners at the Queensland Police Service and have been hand picked from actual case files over several months to challenge even the most experienced examiners.

This component includes our complete data analytic pipeline: from raw individual .txt files to final R Markdown plots and analyses. The “analysis” folder contains data files (.txt) for 12 simulated novice participants and 12 simulated expert participants with pseudorandomly generated response data—these are used to test out the analysis script, and model the null (e.g., what we’d expect to see in our data due to chance). Also included is a LiveCode data simulation tool for producing simulated data, an R Markdown folder with source code, and an html file documenting all code, notes, plots and results of statistical analyses. A ‘final analysis’ folder will be uploaded here once data collection is completed and the human data have been run through this analytic pipeline.

This component contains the LiveCode source file (version 9.0.2) for presenting the sequences and materials to participants and collecting their responses. LiveCode’s Community Edition, is a free open source object-oriented, cross-platform, natural language rapid development environment that can be used for programming local (LiveCode) and web-based (HTML5) experiments. LiveCode is based on the Transcript programming language (a high-level xTalk scripting language like HyperCard’s HyperTalk), and is a great stepping stone to learning Javascript and Python. It is fit for developing experiments that require fine-grained control, complex sequencing, and heavy back-end processing. You can download LiveCode for PC, Linux or Mac at: https://downloads.livecode.com/livecode. You can run this experiment on your own computer by installing the appropriate version of LiveCode Community, and downloading the zip file. Note, however, that the “images” file is missing, which you’ll need for the experiment to run properly. Please contact Jason Tangen (uqjtange@uq.edu.au), and he’ll send you a link to the images we used in this project. Once you get the “images” folder, just put it inside the main experiment folder, which can be stored anywhere on your machine. But the file structure within this folder needs to be preserved to run the experiment, as the script is reading in the participant sequences, images, and instructions on the fly from the same location as the source file.

This component contains the raw individual .txt files outputted and time stamped after completing the experiment.

Preprints, publications, presentations, and other resources related to the communication of this project will be stored in this component.

This component includes our complete data analytic pipeline: from raw individual .txt files to final R Markdown plots and analyses. The "analysis folder" contains data files (.txt) for 12 simulated novice participants and 12 simulated expert participants with pseudorandomly generated response data—these are used to test out the experimental analysis script, and model the null (e.g., what we'd expect to see in our data due to chance; see the simdata.jpeg plots as illustration). Also included is a LiveCode data extraction tool for aggregating raw data from the individual .txt files, an R Markdown folder with source code, and a fully documented html file documenting all code, notes, plots and results of statistical analyses. A 'final analysis' folder will be uploaded here once data collection is completed and the genuine human has data been thread through the data analytic pipeline.

The fingerprint images we used in this project were sourced from the NIST Special Database 300, which consists of 8871 rolled fingerprints, which were donated to NIST by the Federal Bureau of Investigation. Unfortunately, NIST has not made this special database openly available, and we do not have permission to “distribute it further, publish, copy, or disseminate it in any way or form whatsoever, whether for profit or not.” If you’d like to access the images we used in this project, contact fingerprint_data@nist.gov to get permission from NIST to use the database (it’s freely available for biometrics related research, development, and education), forward Jason Tangen the automated access email, and he’ll send you a link to the images we used in this project.

The fingerprint images we used in this project were sourced by our partners in the Queensland Police Services. Unfortunately, these prints are not openly available, and we do not have permission to distribute. If you’d like to access the images we used in this project, please contact Jason Tangen (uqjtange@uq.edu.au).

Fingerprint classification decisions—Is this print from a left or right hand? Is it a thumb, index, middle, ring or little finger?—are central to the upstream fingerprint examination process. Computer algorithms are used to narrow down the search of large national fingerprint databases to a much smaller list of the most highly similar candidate prints. But it is not uncommon for this search process to return dozens of candidate prints for the human examiner to sift through. To help narrow this list even further, fingerprint examiners can nominate which type of finger they think left the latent print: a left or right little, ring, middle, index or thumb. A misclassification of a latent print could wrongly exclude genuine candidates from further examination. Accurate hand and finger classification, on the other hand, helps to exclude candidates that are not a match before the comparison stage, speeding up the process. This project focuses on hand and finger classification as just one particular aspect of the fingerprint examination process. As a first test of hand and finger classification decisions, fingerprint experts and novices are asked to classify prints as belonging to a left or right little, ring, middle, index or thumb. Our goal is to first establish if people can classify prints by hand and finger type above chance. We will then explore whether fingerprint classification performance is a distinguishing feature of visual expertise in fingerprints, setting experienced practitioners apart from novices.

A description of how we generated the images for the experiment can be found in the Wiki below. All of the scrambled images were generated in Photoshop and the actions that we used can be downloaded from the file: "scramble.atn".

In the current project, we are investigating whether there is a memory advantage among forensic fingerprint examiners when recognising fingerprints. In our previous work on memory for fingerprints, Thompson and Tangen (2014) provided preliminary evidence of a superior short-term memory capacity of fingerprint examiners. Unlike our previous investigation of novice vs expert short- and long-term memory, in this current examination of recognition memory, participants are presented with each latent print for 30 seconds, which is plenty of time for uninterrupted analysis and verbal encoding, followed by ten fully rolled prints for them to choose from at their own pace (within a 20 second window). We expect the performance on this task to reflect people’s relative experience with analysing fingerprints.

Preprints, publications, presentations, and other resources related to the communication of this project will be stored in this component.

Preprints, publications, presentations, and other resources related to the communication of this project will be stored in this component.

In the current project, we use a novel keyhole method (Searston & Tangen: OSF Project) for independently sampling parts of fingerprints. We then test fingerprints experts’ and novices' ability to detect prints from the same finger in the complete absence of corresponding features. The keyhole procedure restricts the reliance on corresponding features by sampling small circular patches of ridge detail from separate, non-overlapping locations in each print. To use the more familiar category of faces as an example, imagine only seeing the hint of a smile in one patch of a face, the edge of an eyebrow in another, and then judging whether the two keyholes are from the same person. In this task, it is impossible to rely on correspondence of local features and can only be performed successfully by picking up stylistic information that is distributed across the images. In our first test of this method, we used a discrimination task with pairs of prints presented side-by-side on the screen, to model the latent comparison task that examiners are experience in. Here, we probe the generality of our results across tasks, opting for a one-to-four simultaneous print ‘lineup’ task. We also use more difficult distractor prints, by sampling them from the same location in different fingers of the same person. In this more challenging task, does examiners' experience buy them an ability to see past the pieces to whole print in order to detect matching exemplars.

In this experiment, we will test to see whether expert fingerprint examiners have better visual search abilities compared to novices with a class of stimuli they are familiar with. We will also test whether their superior performance disappears when the structural regularities of the stimuli are removed, or if asked to spot non-diagnostic information. We have conceptualised a fingerprint-like visual search task much like the spot-the-difference tasks you might see in a newspaper to assess how well people can find points of disagreement between prints. Participants will be presented with two fingerprint images side-by-side that are identical but for one fragment that has been replaced with alternative ridge information. They will be asked to spot this difference.

Fingerprint examiners have (many) years of experience dealing with latent fingerprints of varying quality, from partial taps, double taps, inverse and distorted impressions, to pristine fully rolled prints deposited on a ten print card. We would expect then that these experts have become quite good at seeing what a print *would* have looked like had there been no slippage or distortion, or had the finger been fully rolled. We’d expect them to be good at “filling in the blanks” of a latent print, like puzzle, seeing beyond the limits of the pieces in front of them to the bigger picture of the finger. This project aims to test this idea. We've constructed a fill in the blank task, much like a fingerprint puzzle, where we ask experts and novices to identify one of seven fragments that fills in the blank of a print with a hole in it. The fill-in-the-blank task requires an appreciation of information distributed across the print (e.g., ridge flow and thickness) in order to *fill it in* by inferring its' missing features.

Thompson and Tangen (2014) found that expert fingerprint examiners achieved a high level of accuracy in distinguishing prints from the same finger from prints from different fingers even when visual noise was added to the prints obscuring the ridge detail. We are interested in how well examiners can perform under similar “noisy” conditions compared to novices, but when asked to classify fingerprint patterns instead of performing a discrimination task. We know from other experiments we’ve conducted that expert examiners outperform novices in speed and accuracy when detecting, say, an odd loop in array of whorls, so we know that experts’ pattern classification performance already surpasses novices (e.g., Searston & Tangen, 2017). The question now is the extent to which this performance advantage persists when presented with prints that have been superimposed with a high degree of visual noise.