A Communications Perspective on the Use of Visualisations in a Dutch Court for Minor Felonies

Images such as photographs, drawings, animations and videos are everywhere in modern society and are playing an ever-increasing role in court. This general shift from written and spoken language to ‘visual language’ in society has been of interest to language and communication researchers for years. Within the domain of law, however, a lot of the language and communication literature is traditionally focused on written and spoken language. In the first part of this paper we provide a theoretical background in which we discuss how people understand images, what visual literacy means, how images can have persuasive power and can influence people in terms of their emotions, attitudes and decision-making and we talk about how this can play a role in court. Furthermore, we illustrate some gaps in the literature. In the second part of the paper we show, based on a small-scale ethnographic study and systematic observation, how often visualisations are part of the case file in Dutch hearings, what types of images are talked about and/or shown and by whom they are introduced during the hearing. We conclude by discussing what communication scholars find interesting about these observations and we also suggest a number of research opportunities that may be of interest for interdisciplinary research amongst law and communication scholars.


Introduction
On 25 April 2016, in Utrecht, the Netherlands, two suspects are accused by the public prosecutor of demolishing and attempting to steal a moped. The case is handled by one judge at the court of justice in Utrecht, a major city in the Netherlands. 1 The public prosecutor reads her indictment and suggests starting the hearing off with video footage, which, she says, will make it very clear what this case is about. The prosecutor, two suspects, two lawyers and the judge gather around a laptop that is placed at the front of the courtroom. The laptop faces the public; the participants face the laptop. The public prosecutor explains that a moped was placed in a strategic spot by local law enforcement where multiple mopeds had been stolen in the previous months. Law enforcement then waited for a group of youngsters to arrive and filmed what happened. The public prosecutor starts the (night time) video recording and explains who is who, pointing out the two suspects at trial. Sometimes the public prosecutor pauses the video recording and explains what happens. Throughout the video, the judge asks the suspects some questions, including: 'is this you?' One of the suspects replies that he cannot remember if this is him. At the end of the recording the judge states: 'I have seen and recognised you', referring to the two suspects. At the end of the trial, which lasts approximately one hour, the suspects are found guilty of demolishing and attempting to steal the moped. The suspects' presence at the scene is not further discussed during the trial.
The above description, taken from our ethnographic notes, demonstrates a noteworthy phenomenon: the use of video footage in court. Today, images, including video, are everywhere, and we are overwhelmingly exposed to images through television, print and digital (including social) media. Everyone is also able to create images: law enforcement use video (surveillance) cameras to observe the streets (as in the example), bystanders use their mobile phone to film or photograph an event and (semi)professionals create animations or infographics to explain complex structures. At the same time, there are also plenty of tools available that allow anyone to (re)create, modify and share images. The digital revolution, as some scholars call it, has an impact on all sorts of (institutional) domains, including law.
In the example given above we see that the video recording is introduced by the public prosecutor who explains her version of what is happening to the judge. The judge views the video recording live during the trial 2 and asks questions about what he sees in the recording. It is also the judge who decides that he has Lisanne van Weelden, Tessa van Charldorp Utrecht Law Review | Volume 15 | Issue 1, 2019 excluding (audio-)visual aids that facilitate remote communication. 12 Secondly, since we know little about how often images are used in court, we will then report on a small-scale study that gives us insight into how often images are used, when, what types of images they are, by whom they are introduced and whether they are physically shown or not. At the end of this paper we will discuss future directions for research focusing on how (Dutch) court participants possibly process, understand, interpret and describe visual information.

Theoretical background
Before the invention of the photograph (taken by a camera) in 1816, visuals in court consisted of drawings and maps. Around the 1850s, lawyers brought along photographs to court to illustrate situations, and in the 1870s photographs were used to, for example, prove identities of victims or defendants. 13 Almost immediately the use of photographs raised discussions in the field of law: were photographs 'an unmediated replication of nature' or 'manipulable, partial and potentially misleading'? 14 Researchers in the fields of visual culture studies and photography studies have written extensively and critically on the dangers of using images to prove things. 15 Since then many scholars from the fields of law and criminology have conducted various studies concerning the use of visuals in court. Focusing on photographs as well as the use of video and other images, researchers demonstrated whether visuals can raise emotions and have an effect on guilty verdicts 16 and whether or not visuals should be used for identification purposes. 17 We believe that in order to grasp the significance of visuals in court, we have to acknowledge that images can be made by all sorts of people with all sorts of purposes and can be seen and processed by all sorts of other people and interpreted in multiple ways. How meaning is assigned to the visual is hence a very complex matter. Drawing from the Synergetic model of communication, 18 we must see each communication interaction within its context: the sender in the interaction is shaped by individual and societal forces embedded in culture and sends a message through a particular medium within a particular context. At the same time the message is received by a receiver whose individual and societal forces embedded in a particular culture shapes how meaning is assigned to the message. 19 In this paper we argue that first, we need research on the use of various images in court, before we can begin to understand the complex meaning-making processes of images in all its varieties in all sorts of contexts.
In order to do so, we will begin this theoretical section with a definition of 'an image' or 'visualisation' and explore the image as medium within the model of communication. Secondly, we will report on what the current literature says about various communicative functions of images in court and demonstrate that we need more research on this topic. Thirdly, we turn to the receivers of the visuals: we elaborate on what effects visuals can have on emotions and how this may affect the meaning that is assigned to visuals. Fourthly, we explain the notion 'visual literacy' -a concept that is crucial in understanding images on the receiver's end.

Defining the 'image' as a medium
By 'image' we mean any form of visual (re)presentation of objects, persons, events and scenes that conveys a message; we therefore see the image itself as a medium. We also refer to this as 'visuals' or 'visualisations'.
Utrecht Law Review | Volume 15 | Issue 1, 2019 The types of images that typically appear in court are photographs, maps, graphs, diagrams and dynamic images, such as video footage. 20 In the US, these visual presentations are often a representation of the crime at stake. 21 That is, the prosecutor and the defendant's lawyer present their side of the story by reenacting the crime verbally and visually. They tell a (counter)story that explains their case and use images to support its truth-value. The defendant can also use visualisations to undermine the prosecution, who has the burden of proof. In this situation visualisations are used to refute presented evidence.
According to Feigenson and Spiesel, 22 images can be of three different types. The first type includes visual representations that are descriptive of our perception. Hegarty refers to these images as iconic images. 23 Iconic images are representations of objects that themselves are visual-spatial entities. Straightforward examples of descriptive images are photos and film of objects, actions or events. Descriptive images can however also slightly 'distort' reality, for example in a line drawing. A photograph resembles closer how we would have perceived the actual object or event, whereas a line drawing is a translation of our sight and thought. As a result, in these slightly distorted images you will often find some distance between the representation and the actual visual-spatial entity. In street maps, for example, the width of the streets is inaccurate and trees are left out. These two types of descriptive images can have different functions in court. A photograph of a street can be used to indicate the beauty of a tree that should not be removed, whereas a circle on a street map resembling that same tree can be used to show where the tree is precisely located. This shows that descriptive images can be altered or not, depending on the particular task it is to fulfil.
The second type refers to images that are diagrammatic. These are graphic representations that are based on our thought and then translated into a visual representation. Whereas descriptive representations are based on perception, diagrammatic representations are based on cognition. 24 Hegarty refers to these images as relational images. 25 These images represent entities that do not have a spatial referent in reality and therefore are not visible. Examples of diagrammatic images are bar graphs, scatter plots or diagrams, in which visual features like colour, shape and location represent relations between several factors. Such diagrammatic representations are often used in court by scientific experts, who construct a visual to (help) explain facts and opinions.
The third type that Feigenson and Spiesel dissect concerns notational images. 26 These include words or mathematical expressions. Notational images very often form a hybrid visualisation with descriptive and/or diagrammatic images. An example of a combination of the three types of images is a functional magnetic resonance imaging (fMRI) picture. An fMRI scan is a scan of the brain during which the patient performs several tasks, mostly based on cognitive control and attention. The result of the fMRI scan is an fMRI picture showing the slice of a brain in black and white, and coloured areas to indicate areas where the brain was significantly more or less active in the test condition than in the baseline condition. The picture also includes words such as the area in which the active area was found, for example 'prefrontal cortex', or numbers indicating the level of significance. In Feigenson and Spiesel's terminology, the slice of the brain is descriptive, the coloured areas are diagrammatic and words such as 'prefrontal cortex' are notational.
A second interesting distinction between images is whether the image is static or dynamic. Static refers to fixed or still images and dynamic refers to moving images such as film or animation. Research shows that each type has its own cognitive advantages, particularly in learning environments. 27 For example, people can process static images at their own pace and in their own order. This enables them to extract the most relevant elements of the static image and to retain these elements in their memory. The retained information can subsequently be used in other situations. For dynamic images, people cannot control the pace or order of the presentation themselves. However, as dynamic images are assumed to be more realistic, interesting, entertaining and motivating, less cognitive effort might be needed to study the image in the first place.
What the different types of images and their presumed (dis)advantages mean for the use of images in court is until now unclear. At this stage it is even unclear what types of images are actually present in the case file and the hearing. In Section 3, we will report on a study that provides a first insight into this question.

Images and communicative functions of the sender
Images are always created by someone. An important distinction to make however, is that the maker and sender can be (and often are) two different people. In this section we focus on the sender of the images, who uses the image with a particular communicative function in mind. From a communications perspective, images can have multiple communicative functions. Levin has proposed five functions of images that are presented alongside text: they can be used (1) as decoration, making a document look more attractive, (2) to help the reader create a mental representation of an event, person, place or thing, (3) to help the reader to remember the most important information in a text, (4) to help the reader organise information and (5) to explain something to the reader, such as an instruction. 28 Although the text in court is to a large extent spoken, one can easily think of visuals in court with the communicative functions 2, 3, and 4. Yet very little empirical research has been conducted on the communicative functions of images in court.
Some researchers though did make suggestions about the functions that visual material can potentially have in court. Tait reports on a dialogue that occurred between various participants at the conference on graphic and visual representations of evidence and inference in legal settings held in New York in 2006. 29 He states that in the legal process two sorts of images play a major role: images that are used as an instrument of logic and images that are used as a rhetorical device. Although seemingly dichotomous, the two types of images can overlap in particular contexts and should not be perceived as mutually exclusive. Think for example of an aerial photograph in a murder case, on which a murder weapon, a couple of houses and a victim is shown. The photograph shows a relationship between entities and thereby informs on the situation the murder occurred in. Yet the presentational context of the courtroom also makes the image a tool of persuasion. Linguistic descriptions alongside the presentation of the photograph strengthen this persuasive function. Moreover, Feigenson 30 noted that even gruesome crime scene or autopsy photographs, which have been shown experimentally to affect mock jurors' judgments through the emotions they arouse, 31 also convey reliable, probative information about the facts of the case. If this was not the case, the judge should not be admitting them to the hearing in the first place.
Further, Roosma and Dubelaar specifically looked at the Dutch situation and also differentiate between two different communicative functions that visuals can have. 32 First, just like Tait suggests, Roosma and Dubelaar write that visual material in the criminal process can be used to persuade the judge. Second, the authors suggest that visual material can be used to present, visualise, categorise and contextualise evidence. They put the emphasis on the presentation of evidence -which is closely related to the logic function that Tait describes. Roosma and Dubelaar refer to earlier work by Dubelaar and Vanderveen who also distinguished Utrecht Law Review | Volume 15 | Issue 1, 2019 between these two categories. 33 The authors mention that we do not have any information on how often these communicative functions occur, neither do they define what these functions actually mean within the courtroom context.
As we know little about the communicative functions as intended by the senders within court, we suggest that more research is needed to understand this from a sender perspective. There is, however, research to be found on the communicative and cognitive functions various visual types may have from a receiver perspective, which will be explored in the next section.

Receiving the images: vividness and emotion
As described earlier, in the Netherlands the court relies on written statements. This also means that images are described on paper and thus written up as a statement. For example, police officers write up what they saw when looking at a shop's surveillance camera, or describe the photograph made of the victim. Sometimes, the original images are included in the case file, sometimes judges rely solely on the written description of the image. Whereas there may be all sorts of practical reasons for this, whether the original image was actually seen or not may lead to different interpretations of that same image.
Most researchers who consider images as rhetorical devices exemplify gruesome photographs, graphic animations and displays that report scientific findings. 34 The effects that these kinds of images have on people are often connected to experienced vividness and emotion. 35 According to Bell and Loftus, vividness is related to the levels of detail of the presented (in their case) eyewitness testimonies. 36 More detail leads to the experience of vividness, and we are likely to direct more attention to vivid information than pallid information. This increased attention results in a longer and stronger storage in our memory, a good foundation for the rhetorical device to do its work and persuade the viewer. Images are in turn more detailed and vivid than descriptions of those same images and are therefore presumed to be better rhetorical devices than descriptions. Moreover, unlike words, which are constructed by the speaker or writer and are therefore understood as a translation of reality into words, images tend to come across as a representation of the external world without human mediation. 37 As a result of this naive realism, images are almost never questioned for their truth-value.
Images are also assumed to evoke more intense emotions than descriptions of images. IJzermans describes the function of emotions as bridging the gap between the intuitive and deliberate reasoning whilst decision-making. 38 That is, emotions function as spotlights when selecting and processing information. In this context, there seems to be a strong influence of the feeling of (un)certainty that the emotion induces. That is, some emotions are associated with a greater sense of certainty than others; anger, disgust and happiness are emotions that express a feeling of certainty, whereas hope, anxiety and amazement result in a feeling of uncertainty. 39 The greater the sense of certainty, the less inclined people are to process information systematically and the more they are susceptible to heuristic cues. The greater the sense of uncertainty, the more people will try to accomplish the feeling of certainty, by processing information more thoroughly. As such, feelings of uncertainty experienced by a judge can be beneficial for his or her judgment. Judges' emotions in reaction to images can thus have a positive and a negative effect on the decision process.
Different types of images induce these different kinds of emotions amongst the receivers of the images. Bright and Goodman-Delahunty found that mock jurors who saw (gruesome) photographs reported experiencing more intense emotional responses as compared to those who saw no photographs. 40 The photographs mainly induced feelings of anger. Mock jurors that were part of the Douglas et al. study read a detailed trial transcript of a murder trial, with either colour or black and white photographs or no photographs. 41 As compared to the control condition in which no photographs were shown, participants reported more feelings of emotional distress in response to viewing the photographs. Strikingly, the proportion of subsequent guilty verdicts was twice as large as compared to the control condition. These results cannot be translated to the Dutch situation as experienced and highly trained judges interpret these images as opposed to lay people on a jury. We do not know of any empirical research that has carried out such studies with trained judges.
Given Sections 2.2 and 2.3 above, we need to learn more about the role and effects of visualisations in court both from a sender and receiver perspective. What types of visuals are used, with what purpose and how do these visualisations possibly influence the perception and cognition of the participants present during the hearing, especially the judge when dealing with an inquisitorial system?

Receiving the images: visual literacy
One aspect that can have a strong influence on how images are received is one's level of visual literacy. Visual literacy is an indicator of how well we are able to 'read' images and understand them. According to Bamford, being visually literate is a combination of visual syntax and semantics. 42 Visual syntax refers to the pictorial structure and organisation. It concerns the ability to perceive perceptual features such as scale, motion, perspective, background/foreground and draw conclusions from that in terms of camera placement, editing and point of view. Visual semantics refers to the way that we give meaning to the visualisation. When we give meaning to the image we think about who made it, with what purpose it was made, when it was made, whether it was manipulated, what is says about an event, etc. So, visual literacy is a set of skills, not only to interpret the content of visual images, but also to look critically at its purpose, audience and ownership. It is evident that these skills are vital for the participants of the courtroom too. That is, it is important that visual evidence is properly interpreted and that information on the (spatial, temporal, societal, cultural) context of the image is taken into account when meaning is assigned to the image.
Work by Biber on identification of suspects through photographs shows the discrepancy between our perceived visual literacy and our actual visual literacy. 43 That is, our capacity to interpret photos is dependent on photographs that we saw previously. Hence we feel confident; we have seen so many different photographs before. Yet, when we are shown a photograph and asked to identify a suspect based on that photograph, we suddenly experience a strong feeling of doubt. What is it that we see? Are we drawing the right conclusions from the visual? This uncertainty can only be diminished if participants of the court receive training on this type of professional visual literacy.
Next to acquiring visual literacy, participants of the court should also be aware of 'professional vision,' which refers to the way people see and understand events differently based on their professional and cultural background (and consecutive goals). 44 The Rodney King case very clearly shows us that lawyers from different sides can each structure the same piece of visual information (in this case, video footage) according to their own agendas. Goodwin's micro-analysis of both the talk and the image shows how 'talk and visual representation mutually amplify each other'. 45 As such, the talk around the visual makes it even harder to So, with the expected increased amount of images entering the court system and a low professional visual literacy, it is becoming important that participants in the courtroom are being trained to improve their visual literacy. Whilst the majority of law studies (at least in the US and in the Netherlands) strongly focus on written and verbal skills, there might be a need to include visual literacy skills in the near future.

Small-scale ethnography and corpus study
In Section 2 we reported on various studies, some of which have been carried out within the legal domain (often in the US), that show that visualisations can take different forms, serve various communicative functions and have different cognitive effects, each with its own impact on various participants. Before we can further study the effects of visuals in court from a communications perspective, we need more information about what types of images are used, by whom and with what purpose. Only then can we begin to uncover what the effects of the choice of the medium are on its receivers. We therefore set up a smallscale study in the Netherland to find out: 1. What types of visualisations are used in the Dutch court for minor felonies? 2. By whom (public prosecutor or defending party) and how (shown or described) are the visualisations introduced during the court hearing? 3. What relations exist between the factors studied in 1 and 2?
a. Is there a relation between types of visualisations and whether they are shown or described in court? b. Is there a relation between types of visualisations and by whom they are introduced in court?

Method
In this study we combined an ethnographic approach with a systematic observation. During an initial oneweek ethnography study, an observation scheme (see Section 3.2) was constructed. During this ethnography phase, six students and the two authors attended eight hearings, in which we specifically focused on the use of visuals. The ethnographic observations resulted in several criteria to determine the use of visuals. The criteria were tested by the researchers and slightly adapted so that systematic coding could be optimised. During a follow-up two-week period in the spring of 2016, the eight researchers (six students and the authors) observed a total of 85 different criminal law hearings at the Utrecht Court in the Netherlands. The hearings mainly concerned minor felonies. The 85 hearings were coded by hand based on the observation scheme. The coded observations were logged in Excel and SPSS. We performed descriptive statistics and chi-square analyses 46 to analyse the distribution of several characteristics over the observed images in the court hearings.

Operationalisation
The observation scheme requested information on the date, time and the participants present at the hearing. For each image that was presented during the hearing several characteristics were coded. First, it was coded whether the image was literally present or whether one of the participants verbally described an image that was present in the case file but not presented itself at the trial. Second, it was coded which participant introduced the image. Third, the image was characterised as a photo, photo sequence (for example, stills of a video), video, animation, illustration/line drawing, statistical overview or map. Fourth, the types of images were categorised as descriptive, diagrammatic or notational. Fifth, the types of images were categorised as either static or dynamic. Static images included photos, photo sequences, illustrations/line drawings, statistical overviews and maps, whereas videos and animations were categorised as dynamic images.

Results
Our observations, or collection of systematically observed hearings (henceforth: corpus) consists of 85 different criminal law hearings observed at the Utrecht Court in the Netherlands: 80 hearings at the court for minor felonies (one judge) and five hearings at the court for severe felonies (three judges). The mean duration of the hearings in court for minor felonies was 31.5 minutes (SD = 20.3) and for the hearings in the court for severe felonies 99 minutes (SD = 29.67). Table 1 shows the variation of the number of images discussed per hearing. In 48.2% (n = 41) of the hearings, so about half of the cases, images (or descriptions of images) were used, as compared to 51.8% (n = 44) of the hearings in which no images were used. In those 41 hearings in which a visual was used, a total of 75 images was referred to. In most hearings, only one (51.2%) or two (29.3%) images were presented and discussed. Furthermore, our data shows that 72% (n = 54) of the images was actually present in the case file and in the courtroom and shown to the participants, as compared to 28% (n = 21) of the images that was only described by one of the participants. These images may have been present in the case file, but not shown to the participants during the hearing. Our corpus contains for example a description of a photograph showing a battered door after a fight between a suspect and a victim, and a description of a video capturing a car theft.
The majority of the (described or presented) images in our corpus were either photographs or videos. The corpus did not contain animations, illustrations/line drawings or statistical overviews. Hence, all images were categorised as descriptive, rather than diagrammatic. Moreover, about 23% (n = 17) of the images were dynamic (i.e., video) and the remaining 77% (n = 58) consisted of static images (i.e., photographs, photo sequences and maps). Table 2 shows an overview of the types of images found. Over half of the images was produced by law enforcement officials (such as police or local watchmen employed by the municipality). These images were for example photographs of weapons or tools found at the crime scene, photographs of the scene itself (e.g., (suspected) marks of hemp growing pots) and surveillance camera footage of cameras placed at strategic spots. The other images were produced by suspects, victims or, in one case, a witness, see Table 3. Images that for example were taken by victims concerned in five out of seven cases photos of their injuries resulting from abuse. A chi-square analysis shows that there is a significant relation between the type of image and whether the image was shown or described in court, χ(3) = 36.933, p = .000. Fisher's exact test 47 for small sample sizes also shows a significant effect (p = .000). Standardised residuals show that this effect was caused by a higher number of described videos than was expected (n = 14, std. res. = 4.2), a lower number of physically present videos (n = 3, std. res. = -2.6) and a lower number of described photographs (n = 5, std. res. = -2.6), see italics in Table 4. So, during the hearing, videos were often absent and described by participants, whereas photographs were presented. An example of a described video in our corpus concerns footage of a surveillance camera in a supermarket. The public prosecutor describes the video as showing that the suspect is putting several articles in his bag -rather than a shopping basket or cart -and is therefore assumed to be shoplifting. Almost all images (90.7%, n = 68) were introduced during the hearing by the public prosecutor. Only 9.3% (n = 7) of the images were delivered by the defending party. A chi-square analysis shows that there is a significant relation between the type of image and whether the image was introduced by the public prosecutor or the defending party, χ(3) = 10.263, p = .016. However, Fisher's exact test for small sample sizes shows no effect (p = .149). We need larger sample sizes to investigate this relationship.

Conclusion
The results presented above provide insight into how often images are used in a Dutch court for (mainly) minor felonies, what types of image are used, how they were presented, who made the images and by whom they are introduced during the hearing. In about half of the observed hearings images were used. Concerning the types of visualisations that were used (Q1), our analysis shows that mainly photographs (and small number of photo sequences) and videos were employed. As such, the observed images in the present corpus are exclusively descriptive. The absence of diagrammatic images might be explained by the type of court hearings that was attended. First, the hearings concerned cases of criminal law. Other law areas, such as civil cases, might benefit to a larger extent from diagrammatic images. Second, the majority of the attended cases appeared for the court for minor felonies. Minor felonies may less frequently require the presentation of diagrammatic images, for example, visualisations designed by scientific experts to explain relations between facts. Furthermore, Utrecht Law Review | Volume 15 | Issue 1, 2019 about three quarters of the images was static, as compared to one quarter of dynamic images. Although dynamic images are playing an ever-increasing role in today's society, we surprisingly see that video is not (yet) as popular in court as photographs.
When exploring the relation between the types of visualisations and whether they are shown or described in court (Q2 and Q3a), we found that videos were very often implicitly referred to rather than presented to the participants of the hearing. The majority of the photos, on the other hand, were actually shown during the hearing. This finding might also be explained by the fact that most hearings appeared for the court for minor felonies. In this type of court, multiple hearings are combined in a single morning or afternoon session and are characterised by a quick progression of cases. The results showed that the average session lasted about half an hour. In this short time frame, the public prosecutor might find the presentation of videos to take up too much time. Technical possibilities in the courtroom, as well as balancing time investment and seriousness of the case could be reasons why videos are presented or not. Most images we came across were made by law enforcement. However, images from suspects and victims were also introduced and shown. There was only one case in which a witness had supplied an image.
Regarding the parties that introduced the visualisations (Q2), we found that the images were mostly introduced by the public prosecutor. This finding reflects the fact that the public prosecutor is lawfully responsible for assembling the case file 48 and is therefore in charge of the selection and introduction of any visual material present in the case file. Due to the small sample size, our analysis did not show a strong enough relation between visualisation type and introducing party, to draw conclusions from (Q3b).

Discussion
Now that we know that visual materials are used quite regularly in Dutch courts, either explicitly shown during trial (see the example in the introduction of this paper) or implicitly referred to by the judge, prosecutor or lawyer, we need to know more about the effects that visualisations have on the criminal procedure. Earlier reports from the Netherlands suggest that until now there has been very little research on this topic. 49 This is where we believe future research, in the Netherlands and elsewhere, needs to be headed. Below we describe three possible alleys for future research.