A young woman is reported missing. She was last seen in the company of a man who quickly becomes a suspect. Missing Person (MisPer) cases like this are some of the most difficult to solve, mostly owing to the lack of crime scene analysis and forensics (James et al. 2006). In many cases, a suspect is not even clearly identified. In the absence of forensics, investigators are usually left with a “people case”—statements and testimonies acting as the only form of evidence. Researchers have made excellent contributions to help MisPer investigations, from geographical profiling (Keatley et al. 2022; Keatley and O’Donnell 2023) through to statement analysis techniques (Keatley 2023; Richards et al. 2023; Richards and Keatley 2023). Although it is hard to know definite numbers, for a variety of reasons, estimates are that over 600,000 people are reported missing each year in the USA, alone (NamUs 2023). Of course, many are found, and the case is closed without criminal processing. At the other extreme, many MisPer cases are never resolved. This is often not owing to a complete lack of any evidence. Contrary to media portrayals and fictional accounts, cases that involve multiple witnesses, especially if they may be involved in the crime, can lead to more difficulties than potential solutions. Some of the most difficult cases to solve are those wherein the witnesses tell multiple conflicting statements (Keatley 2023; Odinot et al. 2013).

Recently, the author was involved in a case in the USA in which multiple statements were the only evidence and rendered relatively weak owing to apparent inconsistencies. While one person killed the victim, two others were likely involved in the post-mortem activities. Police are given some training to conduct information gathering investigations and accusatorial interrogations to build their cases, but more assistance is required (Keatley 2023). Trace, Interview, Eliminate (TIE) is a fundamental focus of police departments in many countries (Cook 2016). The process of assimilating all of the facts into frameworks is not as clearly trained or regulated. Some officers provide long-form notes and transcripts, others prefer shorthand outlines and summaries. Few officers build tables and databases outlining (in)consistencies between statements, though the use of technology is repeatedly suggested as an important support for large-scale investigations (Keatley and Clarke 2020c). Given the nature of MisPer cases, many “go cold” and the in-depth, perhaps first-hand, expertise and experiences of the lead Detectives or Senior Investigation Officers (SIOs) is lost (Richards et al. 2023). Replacement detectives are often provided with unprocessed raw materials or extensive case notes that are hard to follow or fathom without a great expenditure of time. Decision-making processes are slowed while replacements are brought up to speed on the case.

While police officers have developed ways of timelining narratives and identifying conflicting accounts, more can be done to help highlight key areas. Recently, researchers have highlighted the benefits of using timeline-focused interviews to develop understanding and progress cases (Hope et al. 2013; Punjani et al. 2023). Within the criminology literature, temporal methods are well placed to provide statistical means of analysing timeline statements (Keatley 2018, 2020; Keatley and Clarke 2020a). Many temporal analyses, however, require known ordering of events to compare and contrast timelines. For example, Behaviour Sequence Analysis (BSA; Ellis et al. 2017; Keatley et al. 2017; Taylor et al. 2017), perhaps the most widely used and cited approach, requires clear, detailed temporal timelines. Exact chronological ordering is needed in BSA. Few interviewees can provide this level of detail (Keatley 2018, 2020). This is not necessarily a sign of deception or deflection, as memory fades, especially in cold cases spanning multiple decades.

What is required, therefore, is a simplified method for analysing statements offered by suspects and/or witnesses. Something that not only makes use of current police practice, but positively endorses it, while requiring little to no extra statistical training or data wrangling is beneficial. Furthermore, analyses that are easy to comprehend and read by non-specialist audiences are preferred (Keatley 2020). This is especially true if the outcome of such analyses appears to emerge from “magic box” computations or provide results that cannot easily be interpreted and explained (e.g., to senior officers, district attorneys, and juries). One of the most simplified and effective temporal methods used by criminologists to develop activity timings from vague windows of possibilities is Aoristic Analysis (Andresen and Jenion 2004; Keatley 2020; Ratcliffe 2000, 2002). Aoristic analysis was developed specifically to provide probabilities of criminal activities if individuals could only provide broad time ranges within which a crime had occurred. The typical way of explaining aoristic analysis is through the use of burglaries (or thefts). In these crimes, victims know a crime has occurred, but not exactly when, owing to their absence while the crime was committed. All that victims can typically tell investigators is when they left their house and when they returned. As investigators, let us imagine we are attempting to find the most likely time for burglaries to occur.

Aoristic Interview Analysis

Statements given in police interviews can change. Not every change in a person’s statement reflects an indicator of deception or guilt (Keatley 2023). Some statement analysis approaches suggest coding spontaneous changes of facts to be an indicator of truthfulness (Richards and Keatley 2023; Vrij 2005). The issue, here, is not on decoding truth or deception in a statement, but on highlighting consistencies in a more statistical manner. Indeed, suspects whose statements appear too scripted and exact can often lead to more suspicion than those who are unsure about certain parts (Keatley 2023). Many individuals would struggle to remember precisely what they did last week; remembering events from years prior (as is required in cold case reinvestigations) often leads to self-doubt, variations in timelines, and inconsistencies. What is required is a method that allows investigators to map more clearly where the consistencies and inconsistencies arise. From this, investigators can assign probability weights to certain events occurring at certain times and know exactly where to focus their investigation to clarify timelines.

Aoristic Analysis (Andresen and Jenion 2004; Ratcliffe 2002) has been shown to be a useful approach in outlining probable timings of certain events, where precision is impossible to know (e.g. burglary, theft of unattended belongings). Investigators are well-aware that for an individual case, the precision of the timing may not be exact, but it allows a number of investigative leads to be sought and corroboration prioritised for particular points or areas of the timeline. For example, if a new crime is recorded, investigators may compare the newly recorded crime with the aoristic windows and consider if the crime is connected. Investigators may also compare probabilities between different activities and crimes.

For the sake of ease, imagine 10 victims of burglary are interviewed. Each victim provides an account of only what they know—when they were absent from their home. What we still do not know is when exactly the break-in occurred. Figure 1 provides a graphical display of these data. Each victim tells the investigator what hours they were absent from their house, during which time a burglary occurred, coloured dark grey in the figure. Each row of the graph shows that each person’s account is relatively vague, spanning multiple hours. However, when looking at the columns, we can see a clear time window that is consistent across victims (10–11.30 p.m.). This is the basis of aoristic analysis—investigating the most likely time when a crime occurred. From multiple interview accounts, quantifiable data windows are created, and then probabilities can be calculated (See Ratcliffe 2002 for further details). Of course, it is not a definite pinpointing of each exact time, but it allows investigators to move forward. The suggestion being forwarded in this paper is that the same approach could be used when interviewing suspects/witnesses in complex criminal cases.

Fig. 1
figure 1

Time windows for burglaries. The full aoristic analysis includes aoristic sums and aoristic probabilities. This is outlined clearly in the “Results” section

Full size image

In complex major crimes, such as MisPers and homicides, it is not unusual to have a number of suspects, witnesses, and alibis. The notion that police focus narrowly on one individual is often untrue in real-world investigations. Tip-offs, leads, door-to-door canvassing, and variations in statements ensure police are often tasked with following multiple potential suspects stemming from myriad leads. Within each interview there are typically consistent and inconsistent times given – often stemming from the same individual. In a recent case the author was involved in, for example, an individual had been interviewed several times. Each interview appeared to produce conflicting times leading several investigators to question the overall veracity of the informant. Lining-up the statements on a spreadsheet certainly seemed to highlight how certain activities appeared, disappeared, and changed times. A snap assessment could therefore lead to a conclusion of the individual timewasting or purposefully subverting the investigation. Such decisions should not be made quickly or lightly in real investigations - it may be that the individual is confused or has poor memory. Discounting any individual’s statement(s) can lead to important information being lost. A re-analysis of the statements, using Aoristic Interview Analysis, could help provide some key insights.

The Present Study

This is the first study, to the authors’ knowledge, to use aoristic analysis for interviews. While timeline analysis methods have received multiple supporting publications (Keatley 2020; Marono et al. 2017; Quinn-Evans et al. 2019), including in the area of interviews and statement analysis (Keatley et al. 2018; Richards et al. 2023; Richards and Keatley 2023), none have yet directly tackled inconsistencies in timelines. Path similarity metrics (Keatley and Clarke 2020b) and similar pattern matching algorithms (Abbott and Hrycak 1990; Rosenbaum 1989) may allow investigators to calculate how similar two or more temporal sequence statements are, but it does not highlight exactly when and where consistencies and inconsistencies arise. The current approach, Aoristic Interview Analysis, uses a case study example of how it may be used; however, a note of caution should be understood. The analysis is novel and should not be used to make investigative decisions. The primary goal of this research is to create avenues for further exploration, support, and most importantly development of the approach. It is hoped that academics and practitioners will collaborate on this endeavour, allowing synergistic collaborations to emerge.

Owing to obvious sensitivities of real-world case materials and given this is a proof-of-concept approach, the case outlined here is loosely based on a real-world case that was being investigated and involved multiple accounts across several interviews from people involved. An interesting part of this case was the multiple statements made by a key Witness, which appeared to change. This left investigators in a position of being unsure which narrative was correct and what exactly happened when. Various timelines and matrices were formed in an attempt to build a coherent singular timeline. This is where Aoristic Interview Analysis may help by converting those spreadsheet timeline accounts into probability scores. For reasons of case integrity, identifying features have been removed and certain activities coded purposefully vaguely. Please note, this does not diminish the effectiveness of understanding the Aoristic Interview Analysis process, it simply maintains case integrity. Readers are encouraged to practice the analysis on other cases, perhaps allowing a database of cases to be built.

Methods

Case Summary

The current case involves the abduction and murder of a 10-year-old boy. The boy was last seen leaving a local shopping centre, walking in the direction of his family home, 1 mile away. There was no CCTV footage of the boy leaving the store or his possible route home. The boy, John DOE,Footnote 1 never made it home and his parents reported him missing later that afternoon. Both parents were cleared of any involvement and were not and have never been suspects in the case. Through police investigations, however, a name did arise, and police brought in the initial suspect, “John SUSPECT”, for questioning. Suspect denied any involvement in the abduction of the boy, stating that he was in the presence of “Joe WITNESS” at the time in question. Police then brought in Witness for interview who initially gave an incriminating statement against Suspect. Several days later, while police were building their case, based on Witness’s initial statement, Witness retracted the statement and offered an alternative timeline. The case eventually went cold. Several years later, a new police cold case team re-opened the case and re-interviewed Witness, who appeared to give a third version of the timeline. This left investigators unsure of which statement was more accurate. Several investigators deemed Witness to be unreliable and indicated his statements should be removed from the investigative process.

A third person, John OBSERVER, was identified during the initial investigation as allegedly being in the company of Suspect and Witness. His statements were taken independently across three recorded interviews. Inconsistencies in his statements both internally (his own timelines did not clearly match) and externally (his timelines did not match Suspect or Witness) created further investigative uncertainty.

Coding

The statements offered by the three individuals (Suspect, Witness, and Observer) were recorded and transcribed fully. In all interviews of all individuals, full timelines were taken and discussed. For the purposes of aoristic analysis, timelines were put into chronological order, preserving the temporal chronology of the events as they are alleged to occur, not as they were recounted in the interviews. For example, across an interview, an individual was asked to repeat, clarify, or confirm particular nuances and activities that occurred at particular times. The aoristic analysis takes each of these instances and builds a single, coherent timeline of the events and activities that occurred on the day and times in question. For example, observe the following interaction:

Interviewer: So when did you go to the store?

Suspect: At 2pm

Interviewer: OK, and where were you before?

Suspect: At home

Clearly, a full interview is typically a lot longer and more detailed, but in the current example, the timeline would be “Home-to-Store at 2 pm”. This recodes the individual’s interview statement into a chronological timeline. Of course, people are not always so definite on exact timing of events. As can be seen in Fig. 1, these broader timings can be entered into the aoristic interview analysis. If there were uncertainty about when an event occurred, either because no times were asked or an individual refused to answer, then that action was coded as occurring the entire time between other known actions, for example, if the suspect said they went to work at 9 a.m. and left at 5 p.m. These are the start and end points for being at work. If the suspect indicated, “sometime, during work, I went to the store”, then “store” would be entered as between 9 and 5. This lowers the accuracy of when the event took place but provides investigations with two important outcomes. While 30-min time windows have been used here, aoristic analysis and therefore AIA can have parts of those windows entered. For example, if a person suggests that they left their house at 12.50, then 5/6th of the window would be “at home” and 1/6th would be having left their home (see Ratcliffe (2002) for further explanation). For simplicity, 30-min windows are presented here. At present, there is no guidelines on what the “best” time windows are or should be for timeline analyses (Keatley 2018, 2020). While greater detail often seems to be more descriptive and exact, it can often obfuscate larger temporal trends (Keatley 2020). Therefore, 30-min windows are selected here to provide an overview of the method.

Finally, once interviews had been transcribed and re-ordered into correct chronology, actions and events were coded in accordance with other temporal analysis procedures (Keatley 2018, 2020). Actions and events were given category codes that related to the same event occurring. For example, “I went to Joe Blogg’s store” and “I went to Joe Blogg’s shop” were coded as “went to store”—if both premises were confirmed as being the same place. Similarly, “met with friends” and “met with mates” was coded as “met with friends”. This allows references to the same behaviours to be coded and compared parsimoniously across accounts. In the current case, owing to the depth of investigation, there were no ambiguities arising from word choices/semantics. For note: if such issues do occur, they can either be clarified in follow-up interviews, or else different codes can be used. For example, the difference between “store” and “shop” could be coded as “shopa” and “shopb” to signify that there is a consistency in the type of establishment the individual says they visited, but perhaps an inconsistency in which particular premises they attended. Of course, coding processes and nomenclature can affect probabilities, and so it is important that code books are clearly kept for clarity (Keatley 2020).

Results

Once timelines were chronologically recoded, aoristic graphs were developed for each individual’s interview. Witness’s interview timelines, presented in Aoristic Interview Analysis are shown in Figs. 2, 3, and 4. The Aoristic Interview Analysis process is provided in detailed steps for Witness’s interviews. Observer’s AIA outcome is then provided as analysed to show that it can be computed quickly. A final comparison between interviewees’ statements is then provided.

Fig. 2
figure 2

Windows of activities across each interview for “Witness”. Gridlines left in for ease of interpretation of timings

Full size image
Fig. 3
figure 3

Probability of activities across each window for each interview for “Witness”. Gridlines left in for ease of interpretation of timings

Full size image
Fig. 4
figure 4

Aoristic Interview Analysis for “Witness”. SUM, aoristic sum; gridlines left in for ease of interpretation of timings

Full size image

Witness’s Interview Timelines

For ease of understanding, each activity Witness said they participated in is shaded on the graph. Doing this by colour can provide a quick overview of the overlaps, consistencies, and inconsistencies between each of the three interviews. So, for example, the yellow block activity (which was actually, “I was at home”) was noted as being between 12 and 1 p.m. in the first interview. In the second interview, this was expanded to be between 12 and 1.30 p.m. Finally, by the third interview, it was only noted as being 12–12.30 p.m.

Without any statistical analyses, consistencies and inconsistencies can clearly be seen across the interviews. For example, Witness is consistent at the times he says he was at the Suspect’s house with him—from 3.30 p.m. onwards. Remember that investigators felt the Witness was inconsistent and not reliable. Other consistencies are that being “At park” while slightly varying in times of when this occurred, the duration remained constant (90 min). This descriptive overview is the initial observations of the timings, akin to what many investigators and detectives do with various databases and spreadsheets. In real cases, however, timelines are more complex and involve more activities. To make this analysis more effective, therefore, aoristic interview analysis can be conducted to begin assigning probabilities to each activity.

Activity Counting

Time zones or “windows” are created by the investigators/end-users in aoristic analysis. For ease, 30-min windows were used in the current dataset. Everything from millisecond windows to months, years, or decades could be used. Each activity is then independently given a probability of occurring within each window. For example, in Interview 1, Witness suggested they were at home from 12 to 1 p.m. This offers 2 windows for “At home”, meaning a 0.5 probability of being “At home” in each time window. In Interview 2, “At home” spanned 90 min (3 windows). This meant each cell had 0.33 probability. By Interview 3, Witness suggested they were “At home” only between 12 and 12.30 p.m., so this was given a probability of 1. This process was repeated for all activities stated in each interview, as seen in Fig. 3.

Aoristic Interview Analysis

The final and most important stage of Aoristic Interview Analysis is in bringing together the probabilities of each activity in each window and calculate which activities were more likely to occur at particular time periods. In most traditional types of aoristic analysis, only one activity is analysed (e.g. break-in, burglary, theft). The development for the current research was to combine multiple activities that may overlap or co-occur, across multiple interviews. This makes a more complex analyses but follows the same principles.

Figure 4 shows the AIA for Witness. The row labelled “Conflict” indicates windows in which conflicting activities are stated to have occurred. This is different to standard aoristic analysis, but important in interview settings. This might not always indicate that there is an issue in the statements (sometimes multiple activities may be possible at one point in time). In the current case, the black cells indicate an inconsistency in geographic locations.

The statistical, probability analyses are the most important step of aoristic analysis. The first part is to simply “Aoristic Sum” the probabilities of each activity in each time window. So, the total probability for “At home” from 12 and 12.30 p.m. is 1.83 (M = 0.61). This of course goes above a probability of 1, but it is the aoristic sum probability (Andresen and Jenion 2004). Importantly, this is the highest probability for “At home”, which is only 0.83 (M = 0.28) and 0.33 (M = 0.11) for 12.30–1 p.m. and 1–1.30 p.m., respectively. In pure aoristic analysis terms, this would indicate that being “At home” was most likely at 12–12.30 p.m.—a time that is never contested across the interviews. In aoristic analysis, these sums are then typically divided by the number of sources (Ratcliffe 2002). This is shown as the aoristic probability in Fig. 4. As the day progressed, the probability reduced. The issue here is that if we look at 12.30–1 p.m., then the sum of activities in that time window is difficult to interpret—because different activities are suggested to have occurred in that window. Therefore, in the section below these two rows, probabilities are given for each activity, across the times mentioned. Highest probabilities are given in bold. The initial “Sum” and “Probability” rows are left in, as they relate to standard aoristic analysis computations. The charts beneath these are more important for AIA, as they breakdown time windows per activity.

The AIA means some preliminary probability judgements. For example, “At home” was more likely between 12 and 12.30 p.m. Driving was more likely between 12.30 and 1.30 p.m. “At park” was more likely between 2 and 3 p.m. “At Suspect’s house” was more likely between 3 and 5 p.m. We can also see consistencies within the statement such as duration of events (e.g. at Suspect’s house, at the park) and consistent temporal locations (e.g. at Suspect’s house).

John Observer Statements

Given the nature of AIA, multiple statements can be cross-examined, and probabilities examined separately and later combined. In the first instance, Observer’s interviews were analysed in the same way as Witness’s interviews. The main difference is that Observer suggested he was dropped back at his house after they had all been driving around and to the park.

Figure 5 shows that Observer is more consistent in his statements overall; however, there were still inconsistencies. For example, when exactly Observer was picked up by the Suspect, driven around, and at the park. Also, Observer suggests he was at home at the same time that Witness suggests he was at Suspect’s house with the Suspect. Though they lived very close, so this minor inconsistency could be explained as being dropped off on the way, which supports the travel routes outlined by all individuals.

Fig. 5
figure 5

Aoristic Interview Analysis for “Observer”. Gridlines left in for ease of interpretation of timings

Full size image

Combining Statements

The final additional step of Aoristic Interview Analysis is to bring together statements from different individuals to assess consistencies and inconsistencies between them. In the current case, there are overlaps between the individuals’ statements, which may indicate corroboration. For example, while they vary on when they were driving around, both agree in all interviews that between 1.30 and 2 p.m., they were driving around. In 5/6 (83.3%) of statements, they suggested they arrived at the park, in 100% of statements they agreed they were at the park between 2 and 3 p.m. Only once, in Witness’s second interview did he suggest they stayed at the park until 3.30 p.m.

The next step is to look for conflicting statements with regards timings (see Fig. 6). For example, in Interview 2, Witness suggests they stayed at the park until 3.30 pm. This internally conflicts with his other two statements, and with all 3 of Observer’s statements. Furthermore, in Observer’s second interview, taken at around the same time as Witness’s second statement, he suggested that they actually left the park earlier. This inconsistency may indicate a conflict for further investigation.

Fig. 6
figure 6

Combining statements of Witness and Observer. Gridlines left in for ease of interpretation of timings

Full size image

Discussion

The aim of the current research was to provide a novel approach to analysing multiple interviews for internal and external consistency. Sometimes, in police investigations, it is not the lack of information nor the lack of suspects that impedes progress, but the abundance of conflicting narratives. In some cases, owing to a series of seemingly conflicting statements with varying timelines, investigators may be impeded from building a solid case. Some consultants may suggest that the inconsistent statements should render both individuals’ statements as not viable. The current research shows a new method for analysing multiple timeline statements. Results indicate a possible way to develop probabilistic interpretation of timelines, as well as highlighting consistencies and inconsistencies. The additive/accumulative nature of the research analysis means that as new timelines emerge, they can be added to the analytic framework and probabilities recalculated. This can be expedited via basic computer programs or can be done by hand. While the current research indicates support for aoristic analysis in interview settings, the findings should not be overstated or applied in real-world cases until further support is gathered and statistical calculations agreed upon.

Aoristic analysis has received a lot of support in research across a number of criminal cases (Ashby and Bowers 2013; Ratcliffe and McCullagh 1998). This research adds to that trend and suggests a further step can be taken. Instead of simply looking at one type of crime that occurs at an unknown time across a large group of victims, the analysis could be used to measure consistencies and inconsistencies in suspect and witness statements for a particular crime and surrounding activities. The current manuscript is the first, to the author’s knowledge, to attempt this application of aoristic analysis in a proof-of-concept study.

Research into memory and investigative interviewing techniques could be combined with the AIA approach to develop an understanding of which interview approaches provide more consistently clear timelines. Researchers have highlighted the benefits of using timeline reconstructions in interview settings (Hope et al. 2013). It may be that AIA becomes another approach to improving recall accuracy—by allowing interviewees to see the areas they are inconsistent in recalling. In real-world cases, there may be many years between interviews and the effects of questioning on recall may affect the accuracy of memories (Fisher et al. 2009; Gilbert and Fisher 2006; Odinot et al. 2013). A benefit of the AIA approach is that it may highlight areas that investigators need to focus on to improve accuracy. It may be that features of events affect memory too, such as salient factors like weapon focus (Baddeley 1972; Kramer et al. 1990) or inebriation at the time (Mintzer 2007) or various forms of memory contamination (Paterson et al. 2011). Investigators may attempt to corroborate timings and therefore prioritise certain witnesses’ or suspects’ timelines based on the accuracy or consistency of their statements and independent verification of the facts. It is premature to suggest which interview strategies would work best, but AIA provides a novel approach to begin investigating within a memory research context. Alongside this, an interview that deviates notably from other interviews could be re-examined to look for explanations (e.g. sleep deprivation of the individual, questioning style, contamination).

Police officers are obviously well aware of the need to construct an accurate timeline of activities (Cook 2016). The difficulty is in ascertaining the truth or deception in statements (Keatley 2023; Vrij et al. 2019). While multiple forensic linguistic methods exist, none have proven irrefutable or without criticism (Vrij et al. 2004). A benefit of the current analytical approach is that it does not require pinpoint accuracy of events but could be used to highlight inconsistencies and conflicting statements. In addition, consistencies are also more clearly shown and given a probability value. In the current case, there are multiple consistencies in the Witness’s statement, which may give some support to their overall attempts to recall accurately. This may allow investigators to focus on particular activities or at the least the start and end points of them, in order to narrow down windows.

The Aoristic Interview Analysis also has two other benefits for criminal investigations. First, the statistics are simple enough to be explained by pen and paper. From experience, investigators are more open to using methods if they can fully understand the statistics and process leading to the conclusions (Hackett et al. 2020). Furthermore, many investigators do not have the time or resources to learn and use expensive software and programming. A further benefit is that the Aoristic Interview Analysis provides quick-summary charts that can immediately be used to highlight consistencies and conflicts in statements. This can help to streamline follow-up investigations as well as highlight where to focus follow-up investigations. In the current case, for example, there is no conflict about driving around between 1 and 1.30 p.m. or being in the park between 2 and 3 p.m. This offers some increased certainty about the suspect’s whereabouts during certain times. Returning to interview the key individuals in a case, it is interesting to note whether they attempt to avoid being in these locations at these times. For example, does one person of interest produce different timelines entirely—beyond the scope of probabilities? This is something that requires further exploration and statistical testing.

The current method is not without limitations and, more importantly, key areas for development. First, the analysis should be applied to a wider range of cases with different parameters to check whether it can be applied equally well. For example, how well does it work with only one witness? While it may seem that more interviews and more timeline recitations will give greater accuracy, this may not always be the case. It may be that individuals effectively rehearse their narrative with each retelling. A further area of investigation could be to see whether Aoristic Interview Analysis can be used as a measure of statement change over time. For example, do timelines become more or less exact over time? This could be studied in terms of guilty and innocent individuals. In many cold cases, interviewees have a large gap between interviews. Understanding the length of time between when interviews occur would be an important potential confounding variable. Cognitive psychology researchers specialising in memory and interview analysis could offer great insights into this potential avenue.

Finally, the AIA approach can be adapted to include “known” facts as well as corroborative information and evidence. For example, if bank records, technological mapping/tracing, or forensic evidence can corroborate an individual’s statement timeline, then that individual could be given a greater probability weighting for their statement. When combining statements, for example, instead of giving equal probability weights to each iteration or individual, those times that are forensically shown to be corroborated could be weighted more heavily. Deviations from the accurate timeline may then be explored. It may be that the individual had been threatened or otherwise coerced out of making a statement.

Conclusions

The current manuscript offers an important development of the standard aoristic analysis. Rather than looking at a number of crimes occurring across multiple cases, the analysis has been adapted to ascertain the probability of activities occurring across the commission of a crime, using suspect statements as data. Police are not ignorant to the importance of developing timelines and creating spreadsheets or visual timelines to assist in their investigations. The use of Aoristic Interview Analysis, once more formally tested, developed and supported in the research literature, may assist with investigations. The outcome is an analysis that could help police to ascertain the probability of when certain activities occurred.