Issues in Forensic Psychology No. 8 43

TIME MATTERS in forensic psychology.One of the few things that transcendsthe differences of our subdisciplines is a

concern for understanding and shaping howevents unfold over time. Clinical treatmentsuse a series of interventions to promoteincremental changes in patients’ behaviour.Risk assessments evaluate the trajectory of anoffender’s personal development to predicthis or her future conduct. Police investigationsunfold as a series of decisions and actions.Even serious crimes, the basis of much foren-sic work, comprise a sequence of actions andreactions between the offender and victim.

Given the importance of time in forensicpsychology, we might expect the field to beglutted with analyses of ‘forensic processes’.Unfortunately, the reality is very different. To

examine the extent to which time has fea-tured in forensic research, we reviewed allarticles published in the top 10 forensic psy-chology journals (as identified by Walters,2006) between 2000 and 2007. An article wascounted as analysing process if it considereddata from at least three time points, using amethodology that retained the temporalrelations between these time points. Forexample, a study measuring the impact oftreatment on offender behaviour at delivery,6-month and 12-month follow-up would becounted as measuring process. So would astudy using time-series analysis to examinethe unfolding violent interaction betweenhusband and wife in conflict. In total, wefound just 16 papers that examined processas set out by our criteria.1 This equates to just

Analysing forensic processes: Taking timeinto accountPaul J. Taylor, Karen Jacques, Ellen Giebels, Mark Levine,

Rachel Best, Jan Winter & Gina Rossi

AbstractA great deal of forensic psychology concerns sequences of behaviours or events. In this paper, we review somerecent efforts to examine forensic issues as sequences, discuss some of the contemporary methodologiesinvolved, and highlight some of the lessons that emerge from this research. Specifically, we show: (i) howresearch on public violence has benefited from studying incidents as patterns of cues and responses amongperpetrators and bystanders; (ii) how regularities in the histories of those who undertake suicide terrorismmay be identified by mapping their life events on a graphical timeline; and (iii) how sequence-based corre-lation coefficients make it possible to test detailed theories about the ways perpetrators respond to the vari-ous influence attempts of police negotiators. We conclude by encouraging forensic psychologists toconceptualise their own areas of investigation as a sequence of events rather than a collection of variables.

1 To be counted as analysing process, an article had to consider data from at least three timepoints using a methodology that respected the temporal relations between the three points.Our decision not to include articles using two time points (e.g. the before–after designs preva-lent in recidivism research) was made because such designs speak effectively to what happensbut say less about the process by which it happens. A list of the papers identified as examiningprocess is available at: http://www.lancs.ac.uk/staff/taylorpj/ifp. Please get in touch if youthink we have overlooked an article relevant to process.

Paul J. Taylor et al.

44 Issues in Forensic Psychology No. 8

over two articles a year, and less than 1 percent of all papers published by these journalsover the seven years. By way of comparison,the British Journal of Social Psychology alonepublished seven articles on process withinthe same period.2

The absence of process in forensic publi-cations makes it difficult to draw any conclu-sion except that forensic psychology has notfocused on process nearly as much as itshould have (for a similar conclusion in soci-ology see Abbott, 1990). However, this stateof affairs is arguably not without good rea-son. There continues to be a paucity of avail-able texts on sequence methodologies andtheir application (particularly compared toother methods such as general linear model-ling or qualitative analysis), which makes itdifficult for researchers to identify and useavailable methods. There is also an overrid-ing tendency for researchers to organisetheir studies in terms of variables rather thanevents. A variable-centric approach is famil-iar and, when used appropriately, provides arigorous way of testing theory. But it limitsdrastically what researchers can say about theimportance of history and change to theirphenomenon of interest. For example, thereason for an officer’s decision during amurder enquiry can only be understood inthe context of previous decisions, evidenceseen, and the order in which these decisionsand evidence transpired. It is only by usingsequence methods that we can begin tounderstand this decision because it is onlysequence methods that examine the deci-sion within the context of an unfolding set ofbehaviours and events. In sum, methods thatexamine temporality open up ways ofanswering some of the questions that foren-sic psychologists really want to answer.

In the remainder of this paper we seek toprovide a flavour of what studying process

can offer forensic psychology, and how it hasalready begun to reshape practice. We do soby examining three areas of research. Ourpoint of departure is recent studies of vio-lence in the ‘night-time economy’. Thesestudies demonstrate what can be learnedabout group violence simply by conceptualis-ing an incident as a sequence of actions andreactions, rather than as a collection ofoccurring variables. We then introduce agraphical approach to understandingprocess by examining a sequence of eventsthat occur over a much larger time scale –that of an individual’s life. We review howexamining the life histories of female suicideterrorists can provide important insights intothe critical events, turning points, andprocesses of radicalisation that underlietheir involvement. Finally, we reviewresearch on hostage negotiation, conceptu-alised as a sequence of cues and responses,and how sequence methods have made itpossible to test hypotheses about differencesin the effectiveness of influence strategiesacross cultures. Answers to these hypotheseshave informed negotiation strategy.

One doesn’t have to look farOne of the most striking things to emergefrom studying cases as sequences of events isthat it is not necessary to use sophisticatedanalyses to derive important findings. Often,simply deconstructing an event into asequence of behaviours can be sufficient toilluminate what has previously been over-looked. The recent work on violence in thenight-time economy by Mark Levine and hiscolleagues (Levine et al., 2007) is a com-pelling example of this. Violence and antiso-cial behaviour in public places continues torise in the UK, against a background ofdecline in most crime figures (Nicholas etal., 2007). While alcohol consumption

2 In case of doubt, the number of articles published in the British Journal of Social Psychology(BJSP) is significantly greater than those published in the top 10 forensic journals, as measuredusing a one-way t-test comparing BJSP’s count to the distribution of number of articles pub-lished in the top 10 forensic journals, t(9) = –7.87, p < .01.

Analysing forensic processes: Taking time into account

Issues in Forensic Psychology No. 8 45

undoubtedly plays a part in this phenome-non, it cannot account for the variety ofbehaviour observed among individuals whoengage in night-time economy activities. Thebehaviour of the violent few is modified bothby the social context in which drinking takesplace, and by the cultural traditions thateither inhibit or facilitate aggression in thesecontexts. Understanding when and why vio-lent episodes erupt therefore offers much tocrime reduction initiatives.

To derive an initial understanding ofnight-time violence, Levine et al. (2007) col-lected 42 CCTV clips of violent episodes thatdid not include the intervention of thepolice or private security forces (e.g. bounc-ers). They then coded the observable behav-iour of those involved in the incidents aseither escalating (e.g. punch, kick) or de-escalating acts (e.g. hold back), and notedboth who acted, and to whom their behav-iour was directed. So, for example, a codedbehaviour might be a protagonist actingaggressively towards the target (i.e. protago-nist–escalate–target), or a bystander actingto de-escalate another bystander (i.e.

bystander–de-escalating–bystander). Theythen arranged these actor–action–targetcodes into time order. By analysing theresulting sequences of acts, Levine et al. wereable identify certain features that lead tosevere violence rather than de-escalation,and to tell a compelling story about the pat-terns of behaviour that occur in night-timeeconomy violence.

The first step in examining the night-time violence data was to look at the contin-gencies among the actor–action–targetcodes. In its simplest form, a contingency is acount of the number of occasions that oneevent (e.g. protagonist–escalate–target) isfollowed by a second event (e.g.bystander–de-escalating–bystander) over thesequence being examined.

Table 1 gives an example of contingen-cies in the night-time violence data. The leftpanel of Table 1 shows part of a sequence ofbehaviours and the right panel the countedcontingencies. To derive the counts shown inthe right panel, move from top to bottom ofthe sequence in the left panel and, on eachoccasion, note down the behaviour that

Behaviour Sequence

Contingency Table

Behavioural Response

Prota- gonist Target Bystander

Behavioural Cue

Actor Behaviour Recipient

Esca

late

Ta

rget

Esca

late

By

stan

der

Esca

late

Pr

otag

onis

t

Dee

scal

ate

Byst

ande

r

Dee

scal

ate

Prot

agon

ist

Dee

scal

ate

Targ

et Total

(N)

Protagonist Escalate Target 0 0 1 0 1 2 4

Escalate Bystander 0 0 0 0 0 0 0

Target

Escalate Protagonist 2 0 0 1 0 0 3

Deescalate Bystander 1 0 0 2 1 1 5

Deescalate Protagonist 0 0 0 1 0 2 3

Bystander

Deescalate Target 0 0 2 1 1 0 4

NOTE—To conserve space, not all combination of Actor-Behaviour-Recipient codes are shown in this Table (e.g., Target-Deescalate-Protagonist) since they do not occur in the example sequence.

… Protagonist Escalate Target Bystander Deescalate Protagonist Bystander Deescalate Target Target Escalate Protagonist Bystander Deescalate Bystander Bystander Deescalate Protagonist Bystander Deescalate Bystander Bystander Deescalate Target Bystander Deescalate Bystander Bystander Deescalate Bystander Bystander Deescalate Bystander Protagonist Escalate Target Target Escalate Protagonist Protagonist Escalate Target Bystander Deescalate Target Bystander Deescalate Protagonist Bystander Deescalate Target Target Escalate Protagonist Protagonist Escalate Target Bystander Deescalate Target …

Table 1: An example of a behaviour sequence from the night-time economy data and the resultingtable of contingencies.

Paul J. Taylor et al.

46 Issues in Forensic Psychology No. 8

immediately precedes the behaviourreached.3 The results when tallied and put intabular form should match the table shownin the right-hand side of Table 1. As can beseen from Table 1, some behaviours occurtime and time again regardless of the subse-quent behaviour (i.e. they have a higherbaseline frequency of occurrence). Thismakes it more sensible to examine the pro-portion of occasions when the behaviour isfollowed by a particular second behaviourversus any other behaviour. Popular meth-ods of analysing contingencies tend to basetheir analyses on such proportions (e.g. laganalysis, Bakeman & Gottman, 1997; infor-mation theory, Taylor & Donald, 2003).

In the night-time economy data, an analy-sis of contingencies in this manner hashelped reveal much of the dynamics of theincidents. The first thing that is apparentwhen examining these sequences is thatnight-time economy violence is regulated bythe behaviour of bystanders. Acts of violenceby the protagonist or victim are typically metby interjections by one or more bystanders,and it is the valence of the bystander’sresponses, rather than the behaviour of per-petrators or victims, which is most likely toshape the trajectory of violence. At the sametime, and contrary to the popular stereotype,bystanders are far more likely to try to de-escalate than to escalate violence. Of the 42clips examined by Levine et al., only 4 werecharacterised by bystanders who respondedto perpetrator aggression with more escala-tion than de-escalation. In 33 of the inci-dents, the only response of the bystanderswas de-escalation, no matter how many timesthe perpetrator tried to escalate the situation(the average was 6.8 times).

A second finding to emerge from decom-posing the violent episode in this way relatesto the severity of the incident. In contrast tode-individuation theory (Zimbardo, 1969),which suggests that increasing group sizeleads to a reduction in the controls of per-sonal identity and thus a greater chance ofantisocial behaviour, and in contrast tobystander research (Latané & Darley, 1970),which suggests that increased group sizeleads to greater diffusion of responsibilityand thus less likelihood of bystander inter-vention, the contingency data suggest thatincreasing group size leads to more, not less,regulation. To be precise, as the number ofbystanders involved in the incidentincreases, so too does the amount of escalat-ing behaviour. However, this increase isdwarfed by a far sharper increase in thenumber of de-escalation interventions madeby bystanders, making the ratio betweenescalating and de-escalating behaviour firmlyin favour of de-escalation with this effectbecoming more dominant as group sizeincreases. Group size and social identityinteract in ways that can sometimes lead tothe escalation of violence, but more oftenlead to the de-escalation of violence.

The findings that are beginning toemerge from Levine et al.’s research haveimplications for police understanding andmanagement of night-time violence (cf.Drury et al., 2003). All of the findings pointtowards the importance of the role ofbystanders in regulating violence, suggestingthat efforts to design strategies to combatviolence are likely to benefit from engagingwith the bystander group. The findings sug-gest, for example, that intervention strate-gies should not necessarily seek to remove

3 There are many analogies that can help with understanding the different facets of an eventsequence. One that seems to resonate with many audiences is to imagine capturing data for agame of football. To capture the game dynamics it is necessary to note down, on each pass, theplayer who had the ball and to whom they passed the ball (and, to be exact, something abouthow they passed). Counting the number of passes made, or taking the average number ofpasses per player, tells us very little about the quality of the game and does not readily predictthe final score.

Analysing forensic processes: Taking time into account

Issues in Forensic Psychology No. 8 47

bystanders from the equation but, instead,work to facilitate their regulation of the inci-dent. Similarly, they suggest that messagesencouraging group responsibility and regu-lation may be a valuable part of a widercrime reduction initiative focused on thenight-time economy. Of course, Levine etal.’s preliminary findings open up importantquestions that are best answered by studyingthe unfolding interaction among protago-nists and bystanders. One such questionrelates to the failing of bystander regulationduring incidents in which violence escalatesout of control. Identifying the behaviouraldynamics that characterise these incidents,and comparing these to the dynamics foundin incidents where bystanders successfullyregulate behaviour, is likely to enhance ourunderstanding of the factors that initiate andfacilitate night-time economy violence.

A picture paints a thousand wordsOne of the easiest and most rewarding waysto examine a process is to represent the con-tingencies among its constituent parts in a‘state transition diagram’. These diagramsrepresent events within the process as nodesand contingencies as arrows between the var-ious nodes. The result is a ‘flow chart’ likediagram that allows us to begin to interpretthe ways in which events come together overtime. This interpretation is often made eas-

ier by a shading of the arrows. Typically,thicker arrows are drawn for associationsthat occur far more frequently than wouldbe expected by chance. This kind of method-ology has been used for understanding rapeinteractions (Fossi et al., 2005), the use ofaccommodation in courtrooms (Gnisci,2005), the interaction of husbands and wives(Gottman et al. 1977), and even thesequence of events that leads up to road-traf-fic accidents (Clarke et al., 1999).

One recent application of the methodhas been to understand the process by whichindividuals become involved (and later un-involved) in extremism. Karen Jacques(Jacques & Taylor, 2007) has explored thestructure and development of female suicideterrorists’ life stories, as measured by thesequence of pertinent events that occurredin their lives. Specifically, she has collectedbiographies of female and male suicide ter-rorists from various open sources (e.g. biog-raphies, interviews and martyrdom videos),and coded these biographies for the occur-rence of life events ranging from changingemployment to travelling abroad to partici-pating in jihad. These events can then bearranged chronologically, and the sequencesexamined using a state-transition diagram.By comparing across many sequences, thisdiagram provides a modelling that showssome of the most likely pathways into

Figure 1: State-transition diagram of life events for female suicide terrorists

Paul J. Taylor et al.

48 Issues in Forensic Psychology No. 8

extremism and some of the most likely pre-cursors to committing a violent act.

Figure 1 shows an analysis of the sequenceof events in 23 female extremists’ lives priorto undertaking a suicide attack, as reportedin Jacques and Taylor (2007). The state-tran-sition diagram in Figure 1 represents thefemales’ life histories schematically as arrowsconnecting ‘nodes’ of life events. An arrow isdrawn between two nodes when the eventsassociated with the nodes are found to occurnext to one another in one or more of thefemales’ life histories. The direction of thearrow corresponds with the temporal orderin which the events occurred. For example,since family problems (e.g. arguments,parental violence) sometimes precede havingan affair with another man, an arrow is drawnon Figure 1 from the node representing fam-ily problems to the node representing affair.

To capture the significance of the transi-tions, and to avoid presenting an overly com-plex web of arrows, Figure 1 shades theconnections according to a standardisedconditional probability. A standardised con-ditional probability is a z-score of a condi-tional probability derived using the meanand standard deviation of the sample of con-ditional probabilities that begin with thesame initiating behaviour (i.e. Behaviour Afollowed by any behaviour). The higher theresulting z-score for a contingency, the morecommon that particular sequence of eventswas in the lives of the females. A contingencythat occurred fewer times than average (i.e.standardised conditional probability < 0.00)was removed from Figure 1 (i.e. made invisi-ble) on the basis that this was not a commonpathway and might reflect error within thedata. In contrast, a contingency thatoccurred more than average is representedin Figure 1 by a single line, while standard-ised conditional probabilities above 1.00 (p <.10) and 1.96 (p < .05) are identified withdouble and triple lines, respectively.

An examination of Figure 1 provides anefficient way of evaluating the ‘conceptualmire’ of theories about why people engagein terrorism (Silke, 2001). First, the diagram

suggests that no common pathway of eventsexists for females prior to conducting an actof suicide terrorism. This highlights the com-plexity of the issue and suggests that singlemotive explanations of involvement are notsufficient accounts of what occurs. However,while there is no common pathway to femaleinvolvement, the various pathways shown inFigure 1 share the characteristic of beingpredominantly linear. Events further alongthe sequence do not (re-)connect with previ-ous events in the sequence; no arrows movefrom right to left on the figure. This linearcharacteristic of Figure 1 suggests that it maybe possible to ‘measure’ how far an individ-ual is through the process, which is some-thing that might support risk assessment andthreat prioritisation.

Second, the diagram highlights theimportance of several common key events tothe lives of many females prior to theirattack. Particularly prevalent in Figure 1 isthe variable family problems, which appearsto act as a ‘gate keeper’ to many of the sub-sequent events in the females’ lives. Thus,consistent with findings on other types ofoffender (Payne et al., 2007), an unsettledfamily life appears to lead to changes in cir-cumstances, which opens up – or facilitatesexisting – avenues into terrorism. A com-plete explanation for the central role of thisevent in the female pathways requires repli-cation and further qualitative analysis ofeach person’s life history, but it is possible tomake some general observations. Oneimportant observation is that the centralityof family problems is consistent with recentevidence suggesting that females oftenbecome involved for personal motivations,rather than group or ideological motivationsthat are often associated with male involve-ment (Jacques & Taylor, 2008).

Third, towards the bottom of Figure 1appears a very direct pathway to engagementin a suicide attack, that of marriage followedby having a family member (typically hus-band) killed. This pathway is separated fromthe other events shown in Figure 1, and sug-gests that for some females these two events

Analysing forensic processes: Taking time into account

Issues in Forensic Psychology No. 8 49

represent a clear trigger towards suicide ter-rorism. The pathway to involvement inextremism for these females, and the experi-ences that this encapsulates, is arguably verydifferent from those who engage over alonger sequence of events. Such a strikingpathway towards the end point is quite com-mon on state-transition diagrams of criminallife histories, and might be thought of as afast-track of particular concern to preventionefforts. Conversely, note that other eventswhich we might predict to have a direct asso-ciation with engagement in suicide terrorismdo not appear to have such a direct relation-ship. For example, as can be seen in Figure1, membership of an extremist group (mem-ber group) does not set in motion a clearpathway of events towards committing anattack. This suggests that many females whoengage in suicide terrorism did not engagewith an extremist group in any formal wayover a sustained period of time. Rather, atsome point in this process, they decided toact to support the group’s cause.

On more complex state-transition dia-grams it is possible to find events or behav-iours that assume other roles not found onFigure 1. For example, in an analysis of theprocess of rehabilitation, a backwards linkmay concern an event that causes relapses inpatients’ recovery. Similarly, in an analysis ofinvestigative interviewing, a backward linkmay represent occasions where the inter-viewer returns to a previous line of question-ing. Another type of role concerns events orbehaviours that represent ‘turning points’ inthe overall pathway, with the behaviour thatfollows on from the event either leading to anew set of events further ‘downstream’ orleading back to a set of events encounteredpreviously. The turning-point event thereforerepresents a particularly critical point atwhich to intervene to influence the way a per-son moves through the process. Studies usingstate-transition diagrams will often then selectsuch events for further analysis, in order toidentify why the event is critical and how itmight be possible to influence its relevance.

Dissecting a complex processThere are at least two reasons why state-tran-sition diagrams may not be sufficient for aresearcher. The first is that they do not pro-vide an immediate means of making infer-ences about the differences in process acrosstwo or more subgroups (e.g. across inde-pendent variables). While drawing separatestate-transition diagrams for various sub-groups can prove useful, the visual compari-son they afford tells us less about thelikelihood of a difference in contingencyoccurring by chance. This problem is partic-ularly pertinent when the events of interestoccur rarely; they may not take a prominentposition in the state-transition diagram butthey may represent a critical event in the waythe process unfolds in the subgroups. Thesecond is that state-transition diagrams typi-cally focus on immediate contingencies, andin more complicated processes it is often thecase that behaviours in the sequence arerelated, but only after one or more interme-diate behaviours. For example, a policeinterviewer’s line of persuasive questioningmay lead to an immediate confession by aperpetrator, or it may lead to a confessionfollowing a number of intermittentexchanges. It may thus be important to cap-ture these less direct relationships betweenevents, since it may be these delayed rela-tionships that are critical to how a processunfolds. History may take its time to affectthe present (Taylor & Donald, 2003; Wat-zlawick et al., 1968, p.131).

There are a number of ways to overcomethis problem. One possibility is to adapt con-tingency analysis to be sensitive to this possi-bility by counting behaviours that occur the‘one before last’ (often referred to as lag-2),or the ‘one before last before last’ (lag-3), andso on (Olekalns & Smith, 2000; Taylor & Don-ald, 2003). A second possibility is to examinesequences as a series of larger episodeswherein one can examine the co-occurrenceof behaviours (Taylor, 2002; Taylor & Donald,2004). But neither of these solutions is ideal.Each requires a fairly arbitrary decision abouthow to divide up the data to ensure that rele-

Paul J. Taylor et al.

50 Issues in Forensic Psychology No. 8

vant relationships are captured. Fortunately, arecent development in sequence analysismethods, known as the proximity coefficient(Taylor, 2006; Taylor & Donald, 2007), hasgone a long way to resolving this problembecause it is derived from the intrinsic inter-relationships between events.

The basis of the proximity coefficient isthe notion that events contribute to the samepart of a sequence and have more in com-mon when they occur close together withinthe sequence than when they occur far apart.Events over time occur in ways that are con-nected and organised rather than randomand haphazard (Argyle, 1969; Gottman et al.,1977; Kelley, 1997). This is reflected concep-tually in the proximity approach by associat-ing less immediate relationships in asequence with lower proximity. Specifically,the coefficient equals 0.00 if the events beingconsidered occur only at the beginning andend of the sequence. It equals 1.00 if oneevent immediately precedes the second eventwhenever it occurs in the sequence withoutexception. Values between these two limitsreflect differing amounts of proximitybetween the two events being examined. Theproximity coefficient decreases monotoni-cally as more events are found (on average)to separate the two events being examined(i.e. proximity reduces). It does this in a man-ner that is independent of the length of thesequence, and independent of the numberof times an event occurs in the sequence.

One area where the proximity coefficienthas been valuable is in studying interactionsbetween police and offenders. A good exam-ple of this is in hostage negotiation, wherechains of cues and responses come togetherto form the interaction between police nego-tiator and perpetrator. Ellen Giebels and her

colleagues have used a range of sequencemethods to understand the use and impact ofinfluence strategies within this context(Giebels & Noelanders, 2004). Recent work(Giebels & Taylor, in press a; in press b) hasconsidered differences in influence acrosscultures by comparing the interactionprocesses of negotiations where the perpetra-tor originated from either low-context orhigh-context cultures.4 As with all the exam-ples in this paper, the negotiations were exam-ined by coding the behaviour as it occurred inthe unfolding interaction. Specifically, tran-scripts of the dialogue between police nego-tiator and perpetrator were coded at the levelof utterance using a framework of ten influ-ence tactics (well-known in the field as the‘Table of Ten’, see Giebels & Noelanders,2004), which includes tactics such as beingkind, intimidation and rational persuasion.

To further illustrate the data and how thecoefficients are derived, Table 2 presents anexcerpt of an interaction adapted from anegotiation with a high-context perpetrator.This table is divided in a similar way to Table1. In the left-panel is a sequence of codedutterances as spoken by the police negotia-tor (PN) and the perpetrator (HT). In theright-hand panel is the matrix of proximitycoefficients derived from the behavioursequence in the left panel. When laid out inthis way it is possible to explore how interre-lationships among behaviours are reflectedby different values of the proximity coeffi-cient. For example, instances of the perpe-trator sharing information (i.e. HTinformation sharing) always occur directlyafter a police negotiator expresses empathy(PN empathising). Consequently, the prox-imity of these behaviours is the maximumpossible. In contrast, the police negotiator’s

4 The theoretical and empirical background to the distinction between low- and high-contextperpetrators is best understood by reading the original works of Giebels. Briefly, however, asnoted by Hall (1976), low-context communicators favour the use of direct and explicit mes-sages in which meanings are principally contained in the transmitted messages. In contrast,high-context communicators engage in a less direct way, with information being more hiddenand meaning located in the social context of the interaction.

Analysing forensic processes: Taking time into account

Issues in Forensic Psychology No. 8 51

persuasive arguments (PN persuasive argu-ments) occur towards the beginning of thesequence, while the hostage taker’s compro-mise (HT compromise) occurs at the end,which means that the proximity of thesebehaviours is low. Consistent with theseexamples, the coefficient matrix reports aperfect association between PN empathisingand HT information sharing (1.00) and alow proximity between PN persuasive argu-ments and HT compromise (.22). All of theother relationships in the example sequencefall between these two extremes and, accord-ingly, have coefficient values that depend ontheir distances apart in the sequence.

A few other aspects of the matrix in Table2 are worth noting because they illustrate howthe proximity approach helps examineprocess. One is the occurrence of missingcoefficients (e.g. for PN empathising to HTpersuasive argument). These reflect the dis-tribution of behaviours in the negotiation,with missing values in a variable row (or col-umn) indicating that most observations of thebehaviour occur toward the end (or begin-ning) of the interaction (as is the case forcompromise in Table 2). A second is the coef-ficients that appear on the diagonal of thematrix. These provide a measure of reciproc-

ity (Putnam & Jones, 1982) that quantifies thenumber of codes that occur before reciproca-tion, rather than simply the occurrence ofimmediate reciprocation. A third is the asym-metric nature of the coefficient values. Forexample, the coefficient for PN empathisingpreceding HT threats (i.e. .89) is higher thanthe coefficient for PN threats preceding HTempathising (i.e. .56). In general, proximitycoefficients will be asymmetrical, reflectingthe possibility that one event occurs beforethe second on the majority of occasions. Insuch cases, the difference between two coeffi-cient values may provide an indication ofdominance between the behaviours.

One of the advantages of the proximitycoefficient is that it is possible to derive amatrix of coefficients – similar to that inTable 2 – for each sequence in the data set.This is useful because it becomes possible toderive coefficients for the samecue–response relationship across a numberof cases. This, in turn, makes it possible toboth derive means and standard deviationsfor the coefficients, and make comparisonsacross subgroups of coefficients using stan-dard tests (e.g. ANOVA). In their hostagenegotiation analysis, Giebels and Taylor (inpress a) used this approach to examine dis-

Hostage Taker’s Response

Police Negotiator’s Cue

Pers

uasi

ve

argu

men

ts

Info

rmat

ion

shar

ing

Thre

ats

Empa

this

ing

Com

prom

ise

Persuasive argument 94 92 82 33 22

Information sharing -- -- -- -- 100

Threat -- 89 100 56 44

Empathising -- 100 89 78 67

Compromise -- -- -- -- --

… PN Persuasive Argument HT Threat PN Persuasive Argument HT Persuasive Argument PN Persuasive Argument HT Information sharing PN Empathising HT Information sharing PN Threats HT Threats PN Persuasive Argument HT Information sharing PN Empathising HT Information sharing PN Empathising HT Information sharing PN Empathising HT Empathising PN Information sharing HT Compromise …

Table 2: An example of a hostage negotiation sequence and the resulting proximity coefficient matrices

Paul J. Taylor et al.

52 Issues in Forensic Psychology No. 8

tributions of coefficients calculated fromincidents involving perpetrators from high-and low-context cultures.5 The results of theircomparisons across cultural groups begin tooffer insights for those interacting with per-petrators from low- or high-context cultures.

The first lesson to emerge from Giebelsand Taylor’s research is that rational persua-sion, often considered the backbone ofnegotiation strategy, is typically more effec-tive with low-context than high-context per-petrators. For example, a comparison ofresponses to persuasive argument suggeststhat low-context perpetrators are moreimmediate to respond with compromisesthan their high-context counterparts (prox-imity coefficient = .86 compared to .65).Interestingly, this finding emerged irrespec-tive of whether analysis considered interac-tion from early ‘crisis’ periods (so-calledbecause of their association with high-anxi-ety, emotionality, and cognitive deteriora-tion) or later ‘normative’ periods, where theextremes of the context have dissipated. Per-haps because logic and deductive thinkingare valued in low-context cultures (Gelfand& Dyer, 2000), arguments elicited compro-mises from low-context negotiators regard-less of whether or not they were in crisis.

The second lesson to emerge from thisresearch is that using threats to influenceinteraction often results in an escalation ofconflict with high-context perpetrators.Specifically, while low-context perpetratorsare found to be more likely to communicatethreats, high-context perpetrators were morelikely to reciprocate them. One compellingexplanation for this finding, which carriesimportant lessons for practice, is that threats

draw attention to the need to preserve face,and that this need is sufficiently importantwithin high-context cultures for them toreciprocate the explicit challenge (Ting-Toomey & Oetzel, 2001). Threats are a con-frontational way of handling an interactionand, while this is consistent with low-contextinteractions, it is generally more inappropri-ate in high-context cultures (Fu & Yukl,2000). Thus, it is not surprising that threatsare used more by low-context perpetrators,but that high-context perpetrators are morelikely to respond to them by ‘punishing’ withcounter-threats. This is likely to be particu-larly true in high-stakes criminal settings (e.g.police interviews) where establishing domi-nance and saving face may become particu-larly salient factors (Adair & Brett, 2004).

DiscussionAnalysing process holds great promise forforensic psychology. Most if not all of ourclassic theories and contemporary issuesinvolve development or change over time.Our practices also comprise events that aresequential or interactional rather than staticand variable-centric in nature. It is time ouranalyses reflected this. The methods we con-sider in this paper make it possible toaddress forensic issues as sequences. Whatthe field needs are scholars and practitionerscommitted to advancing existing wisdom byuncovering the sequential regularities thatlie behind forensic phenomena. These pio-neers will adopt cases not variables as thefocus of their enquiries. They will talk aboutdevelopment and change, turning pointsand critical pathways, life courses andphases, and facilitators and inhibitors.

5 In Giebels and Taylor (in press a), the comparisons of proximity coefficients are taken onestep further by the use of randomisation tests. A randomisation (or permutation) test assessesstatistical significance in a way comparable to conventional parametric tests, but rather thancompare the resulting test statistic (e.g. an F value) to the critical values found in the back ofstatistics textbooks, it compares it to a distribution obtained from re-sampling of the availabledata (see Edgington, 1995). The advantage of this approach is that it avoids many of the dataassumption of traditional significance tests (Dunlap et al., 2003) and so provides a reliable wayto derive significance levels for the proximity coefficients.

Analysing forensic processes: Taking time into account

Issues in Forensic Psychology No. 8 53

There are of course many forensic andinvestigative processes that we have not con-sidered (for an overview see Donald & Tay-lor, 2007). One core area for investigativepsychology is the problem of differentiatingoffenders on the basis of their behaviour atthe crime scene (Canter, 2000; Nutch &Bloombaum, 1968). The intriguing aspect ofthis research is that, overwhelmingly, it hasfocused on the behaviour of the offender(rather than the interaction betweenoffender and victim) using a methodologythat takes absolutely no account of the orderin which behaviours occur during the crime.Not only does this approach neglect theimportance of interaction process, but it is atodds with what we know from interactionisttheories of human behaviour (Mischel,2004), which is that people do not act con-sistently in different situations. As othershave acknowledged (Alison et al., 2002), wemust assume either that victims provide afairly uniform context (something shown tobe false, see Ullman, 2007) or that the mod-els are robust enough to withstand context-based changes in behaviour. It would seemimportant, therefore, to test whether offend-ers’ behaviour is consistent when faced withchanging behaviour from their victims. Or, ifthis is not the case, to ensure that existingmodels provide a robust measure of inter-personal style from which to develop infer-ences about offenders’ characteristics.

To test this possibility, Jan Winter isundertaking doctoral research that exam-ines whether or not existing typologies ofcriminal behaviour can capture actual con-sistencies in the way in which an offenderbehaves as observed in an unfolding interac-tion between offender and victim (Winter &Rossi, submitted; Winter & Taylor, 2006a,2006b; see also Fossi et al., 2005). To achievethis, Winter is examining victims’ accountsof their sexual assaults and piecing togetherfrom these accounts the sequence ofoffender and victim behaviours as itoccurred during the offence. The result isevent sequences that capture the interactionbetween offender and victim, which should

allow for an analysis of the interpersonalstyle of the offender. Winter’s initial analysesreveal some important lessons for those inthe business of making inferences fromcrime scene behaviour. First, when offendersare examined according to the similaritiesand differences in their interpersonal style(i.e. the ways in which they respond to theirvictim’s behaviour), it is not possible to repli-cate the typologies found previously whenexamining offender behaviour alone. Sec-ond, when an offender’s interpersonal styleis viewed as a function of a number of con-text variables, then consistencies and differ-ences in behaviour begin to emerge in aform that is amenable to theoreticallygrounded classification. Thus, if this prelim-inary analysis stands up to further scrutiny, itsuggests that offender differentiation shouldbegin to focus on interaction processes andthe context in which an offender behaves.

While the examples in this paper havefocused on investigative issues, it is a mistaketo assume that process is not equally impor-tant to the broader church of forensic psy-chology. Over the last five years there havebeen isolated but impressive pieces of processresearch that have illustrated just what can belearned from studying forensic psychology asa process. For example, in the courtroom,Gnisci (2005; Gnisci & Bakeman, 2007) hasshown the value of examining hostile exami-nations in a criminal case as question–answerexchanges. Among other things, his findingsdemonstrate ways in which lawyers use verbaland non-verbal behaviours to control a wit-ness’s narrative freedom and their ability tomake the point they wish to make. In a verydifferent area, the recent study by Warehamand Dembo (2007) used latent growth mod-els to examine the changes in psychologicalfunctioning of juveniles arrested on misde-meanour or felony charges. They show howmodelling change in psychological function-ing over time can provide not only a morereliable indication of the factors that influ-ence development, but also an assessment ofwhen (i.e. at what point on the process) totarget intervention and prevention.

Paul J. Taylor et al.

54 Issues in Forensic Psychology No. 8

ConclusionCriminal acts and investigative decisionsoccur not as variables to be counted but asevents to be understood within a largersequence of events. This is often acknowl-edged in forensic theory, but rarely exam-ined in forensic research. As we haveexplored, there are simple and more com-plex methods available for examining foren-sic processes that correspond well with theway the issue occurs in practice. It is, webelieve, risky for researchers to draw conclu-sions about process when they use a method-ology that does not examine process. As wehave explored in this paper, there are arange of simple and more complex methodsfor examining such sequences and it is timefor forensic psychologists to take advantageof these methods. We therefore encourageresearch and practitioners alike to take thetime to understand and apply sequencemethods in their own areas of work. To assistthose willing to take up the challenge, we

have made available a number of resourcesat: http://www.lancs.ac.uk/staff/taylorpj/ifp.htm. Among other things, you can down-load from this website details of the materialdiscussed in this article, a biography offorensic research that examines ‘processdata’, and software to examine process inyour own data.

AcknowledgementThe research we report was made possible byESRC grants PTA-026-27-0044, PTA-030-2006-00351, and RES-000-23-1098 (awardedto Taylor, Jacques, and Levine, respectively),a British Council-NWO partnership in sci-ence grant (PS844), awarded to Giebels andTaylor, and an OZR grant (#1589) awardedto Winter and Rossi.

Address for correspondenceDr Paul J. Taylor Department of Psychology, Fylde College,Lancaster University, Lancaster LA1 4YF.

ReferencesAbbott, A. (1990). A primer on sequence methods.

Organization Science, 4, 375–392.Adair, W.L. & Brett, J.M. (2004). Culture and negotiation

processes. In M.J. Gelfand & J.M. Brett (Eds.), Thehandbook of negotiation and culture (pp. 158-176).Palo Alto, CA: Stanford University Press.

Alison, L., Bennell, C., Mokros, A. & Ormerod, D.(2002). The personality paradox in offender pro-filing: A theoretical review of the processesinvolved in deriving background characteristicsfrom crime scene actions. Psychology, Public Policy& Law, 8, 115–135.

Argyle, M. (1969). Social interaction. New York: Ather-ton Press.

Bakeman, R. & Gottman, J.M. (1997). Observing interac-tion: An introduction to sequential analysis (2nd edn).New York: Cambridge University Press.

Canter, D.V. (2000). Offender profiling and criminaldifferentiation. Legal and Criminological Psychology,5, 23–46.

Clarke, D.D., Forsyth, R. & Wright, R. (1999). Junc-tion road accidents during cross-flow turns: Asequence analysis of police case files. AccidentAnalysis and Prevention, 31, 31–43.

Donald, I.J. & Taylor, P.J. (2007). Investigative psy-chology. In D. Clark (Ed.), Encyclopedia of law andsociety: American and global perspectives (pp.

821–824). New York: Sage.Drury, J., Stott, C. & Farsides, T. (2003). The role of

police perceptions and practices in the develop-ment of ‘public disorder’. Journal of Applied SocialPsychology, 33, 1480–1500.

Dunlap, W.P., Burke, M.J. & Smith-Crowe, K. (2003).Accurate tests of statistical significance for r-sub(WG) and average deviation interrater agree-ment indexes. Journal of Applied Psychology, 88,356–362.

Edgington, E.S. (1995). Randomization tests. New York:Marcel Deckker.

Fossi, J., Clarke, D.D. & Lawrence, C. (2005). Bed-room rape: Sequences of sexual behavior instranger assaults. Journal of Interpersonal Violence,20, 1444–1466.

Fu, P.P. & Yukl, G. (2000). Perceived effectiveness ofinfluence tactics in the United States and China.Leadership Quarterly, 11, 251–266.

Gelfand, M. & Dyer, N. (2000). A cultural perspectiveon negotiation: Progress, pitfalls, and prospects.Applied Psychology: An International Review, 49,62–99.

Giebels, E. & Noelanders, S. (2004). Crisis negotiations:A multiparty perspective. Veenendaal, The Nether-lands: Universal Press.

Giebels, E. & Taylor, P. J. (in press a). Interaction pat-

Analysing forensic processes: Taking time into account

Issues in Forensic Psychology No. 8 55

terns in crisis negotiations: Persuasive argumentsand cultural differences. Journal of Applied Psychol-ogy.

Giebels, E. & Taylor, P.J. (in press b). Interaction pat-terns of social influence in crisis negotiation. InR. Rogan & F. Lanceley (Eds.), Contemporary theoryand research in crisis negotiations. Westport, CT:Praeger.

Gottman, J.M., Markman, H. & Notarius, C. (1977).The topography of marital conflict: A sequentialanalysis of verbal and non-verbal behavior. Journalof Marriage and the Family, 39, 461–477.

Gnisci, A. (2005). Sequential strategies of accommo-dation: A new method in courtroom. British Jour-nal of Social Psychology, 4, 621–643.

Gnisci, A. & Bakeman, R. (2007). Sequential accom-modation of turn taking and turn length: A studyof courtroom interaction. Journal of Language andSocial Psychology, 26, 234–259.

Hall, E.T. (1976). Beyond culture. Garden City, NY:Anchor Press.

Jacques, K. & Taylor P J. (September, 2007). Pathwaysto suicide terrorism: Do women follow in men’sfootsteps? Poster presented at the BPS Division ofSocial Psychology conference. Kent.

Jacques, K. & Taylor, P. J. (2008). Male and female sui-cide bombers: Different sexes, different reasons?Studies in Conflict and Terrorism, 31, 304–326.

Kelley, H.H. (1997). Expanding the analysis of socialorientations by reference to the sequential-tem-poral structure of situations. European Journal ofSocial Psychology, 27, 373–404.

Latané, B. & Darley, J.M. (1970). The unresponsivebystander: Why doesn’t he help? New York: MeredithCorporation.

Levine, M., Best, R. & Taylor, P J. (September, 2007).Bystanders and the informal regulation of vio-lence in the night-time economy. Presentationgiven at the BPS Division of Social Psychologyconference. Kent, UK.

Mischel, W. (2004). Toward an integrative science ofthe person. Annual Review of Psychology, 55, 1–22.

Nicholas, S., Kershaw, C. & Walker, A.(2007). Crimein England and Wales 2006/07. Home Office Sta-tistical Bulletin 11/07. London: Home Office.

Nutch, F.J. & Bloombaum, M. (1968). A smallest spaceanalysis of gang boy behaviors. Pacific SociologicalReview, 11, 116–122.

Olekalns, M. & Smith, P. (2000). Understanding opti-mal outcomes: The role of strategy sequences incompetitive negotiations. Human CommunicationResearch, 26, 527–557.

Payne, B.K., Gainey, R.R. & Carey, C.S. (2007). All inthe family: Gender, family crimes, and later crim-inality. Women & Criminal Justice, 16, 73–89.

Putnam, L.L. & Jones, T.S. (1982). Reciprocity innegotiations: An analysis of bargaining interac-tion. Communication Monographs, 49, 171–191.

Silke, A. (2001). The devil you know: Continuing

problems with research on terrorism. Terrorismand Political Violence, 13, 1–14.

Taylor, P.J. (2002). A cylindrical model of communi-cation behavior in crisis negotiations. HumanCommunication Research, 28, 7–48.

Taylor, P.J. (2006). Proximity coefficients as a measureof interrelationships in sequences of behavior.Behavioral Research Methods, 38, 42–50.

Taylor, P.J. & Donald, I.J. (2003). Foundations and evi-dence for an interaction based approach to con-flict. International Journal of Conflict Management,14, 213–232.

Taylor, P.J. & Donald, I.J. (2004). The structure ofcommunication behavior in simulated and actualcrisis negotiations. Human CommunicationResearch, 30, 443–478.

Taylor, P.J. & Donald, I.J. (2007). Testing the relation-ship between local cue–response patterns andglobal dimensions of communication behavior.British Journal of Social Psychology, 46, 273–298.

Ting-Toomey, S. & Oetzel, J.G. (2001). Managing inter-cultural conflict effectively. Thousand Oaks, CA: Sage.

Ullman, S. (2007). A 10-year update of ‘review and cri-tique of empirical studies of rape avoidance’.Criminal Justice and Behavior, 34, 411–429.

Walters, G.D. (2006). Measuring the utility of journalsin the crime-psychology field: Beyond the impactfactor. Journal of the American Society for InformationScience and Technology, 57, 1804–1813.

Wareham, J. & Dembo, R. (2007). A longitudinalstudy of psychological functioning among juve-nile offenders: A latent growth model analysis.Criminal Justice and Behavior, 34, 1241–1261.

Watzlawick, P., Beavin, J.H. & Jackson, D.D. (1968).Pragmatics of human communication: A study of inter-actional patterns, pathologies, and paradoxes. Lon-don: Faber.

Winter J. & Rossi, G. (submitted). Integrating interac-tion: Capturing the dynamics of stranger sexualassault.

Winter, J. & Taylor, P.J. (June 2006a). Towards a typol-ogy of crime scene interaction. Presentation atthe Sixteenth Conference of the European Asso-ciation of Psychology and Law, Liverpool.

Winter, J. & Taylor, P.J. (July 2006b). The importanceof the offender-victim interaction in rape. Presen-tation given at the Second International Summerconference: Research in Forensic Psychiatry.Regensburg, Germany.

Zimbardo, P.G. (1969). The human choice: Individu-ation, reason, and order vs. deindividuation,impulse and chaos. In W. J. Arnold & D. Levine(Eds.), Nebraska Symposium on Motivation (Vol. 17,pp.237–307). Lincoln, NE: University ofNebraska Press.