One and a half centuries of serial homicide in Italy: offender, victim and offence characteristics

Abstract

The present study reports offender, victim and offence characteristics of the entire population of known serial homicide offenders identified in Italy between 1848 and 2019 (59 offenders, 244 victims, of whom 50.4% were men, and 21.7% were sex workers). We found that most of the offenders (72.4%) had a personality disorder. The offenders’ age during their series was 35.1 (SD = 11.3) years, on average. Sexual elements were found in 28.4% of the crime scenes. The median time interval between homicides was 2.8 months. Also, we investigated the consistency of the offence characteristics over the series and found significant correlations between the offence characteristics between a previous and the subsequent homicide.

Introduction

Even if serial homicide (SH) is a topic that has captured the attention of the entertainment industry more than that of the scientific community, the phenomenon has received increasing research attention in the last decades. The empirical studies published have mainly focused on three issues: (a) offender profiling and classifying types and motives of SH (Harbort & Mokros, 2001; Holmes & DeBurger, 1985; Holmes & Holmes, 1998; Keppel & Walter, 1999; FBI, 2008; Ressler & Burgess, 1985; Ressler et al., 1986; Sturup, 2018); (b) studying the feasibility of behavioral crime linking (Pakkanen et al., 2021; Pakkanen et al., 2012; Salo et al., 2013; Santtila et al., 2008); (c) analyzing offender spatial behavior (Hickey, 2015; Holmes et al., 1998; Lundrigan & Canter, 2001a, 2001b; Snook et al., 2005; Synnott et al., 2019).

Offender profiling and types and motives of SH

It is often believed (see, for example, Miller, 2014) that the FBI’s well-known classification scheme in organized and disorganized murders is a way to classify SH offenders. However, it should be noted that Ressler and Burgess (1985) conducted their research on 25 SH offenders and 11 sex homicide offenders who had committed a single homicide, a double homicide or a spree homicide. Therefore, their findings are not just aimed at classifying SH offenders.

Ressler et al. (1986) proposed that offenders’ personality and behavior characteristics are like a fingerprint that an offender leaves at a crime scene. This fingerprint refers to two distinct types of offender: organized offender and disorganized offender. According to their model, which is often associated with the SH issue, an organized offender (average to above-average intelligence, socially competent, skilled work preferred, sexually competent, controlled mood during crime, living with partner) leaves an organized crime scene (planned offence, restraint used, body hidden, weapon/evidence absent); a disorganized offender (below-average intelligence, socially inadequate, unskilled work, sexually incompetent, anxious mood during crime, living alone) leaves a disorganized crime scene (spontaneous offence, minimal used of restraints, body left in view, weapon/evidence left in view; see Ressler & Burgess, 1985, for a complete description of the offender characteristics and the offence characteristics). Their classification could therefore be used also as an investigative tool to make offender profiling of the unknown offender through crime scene analysis a means of supporting investigators in criminal investigations. Despite its popularity, when this classification was empirically tested on a sample of 100 U.S. SH offenders it received little support. Canter et al. (2004), using a multidimensional scaling procedure that examined the co-occurrence of 39 offence characteristics with every other, revealed that there are no discrete sub-sets of offence characteristics that can be regarded as distinctly related to the organization or disorganization of the homicide. However, Mjanes et al. (2017), examining a sample of 350 Canadian cases of sexual homicide, found that sexual homicide can be separated into two distinct profiles that share similarities with the organized/disorganized dichotomy in terms of the detection avoidance strategies, control and type of violence used by the offender.

Regarding the use of offender profiling in SH, to the best of our knowledge, there are no empirical studies published in peer review journals on the feasibility of using offender profiling in SH. All the empirical studies published on offender profiling have been conducted on a single homicide sample (see Fox & Farrington, 2018).

Holmes and Holmes (1998), on the basis of the previous work developed by Holmes and DeBurger (1985), propose a classification of SH offenders in four categories: visionary (who kill in order to obey sorts of supernatural entities like God, Satan and angels who command to the offenders to kill someone); missionary (who feel the mission to kill to rid the world of a group of people who are judged to be unworthy or undesirable); hedonistic killer (subdivided into lust killer, who kills for sexual gratification, and thrill killer, who kills for the pleasure and excitement of killing; and power/control (the motive of their homicides relies on their need for power and dominance over another human being). Despite the popularity of this classification, Canter and Wentcik (2004) found that the category of ‘power or control’ killings were found to be common in the sample as a whole and thus did not form a distinct type. Authors found a limited support for aspects of the lust, thrill and mission styles of killing.

Another well-known classification of SH was made by Keppel and Walter (1999) who classified serial sexual homicide as: power-assertive (their motive is not to kill but the escalation in aggression in order to control the victim results in her eventual death); power-reassurance (plans to rape the victim but the homicide can occur when the rape fails due to the offender’s impotence and because of that he attempts to reassure himself of his power and control, killing the victim); anger-retaliatory (the rape and the homicide are driven by the offender’s anger and hatred of women); and anger-excitation (the main aim of these offenders is the pain and suffering of the victim, from which the offender derives sexual pleasure and satisfaction). This classification was subjected to empirical verification by Bennell et al. (2013) who found no evidence of highly co-occurring behaviors/characteristics from Keppel and Walter’s classification, indicating that the classification system is potentially invalid.

A few SH descriptive studies were conducted in the United States.

In a non-peer-review study, Hickey (2015) reported the offender and victim characteristics of the of 431 SH offenders (83% male) over a long period of time in the United States. between 1800 and 2004. He found that the mean offender’s age at the onset of their series was 28 years, that 63% were previously convicted, and the majority of SH offenders attained a high school or less and had a low-level employment. Hickey (2015) reported that 83% of the victims were male, and 61% of SH offenders murdered strangers who were often young females (sex worker, transient or drug addict).

Pakhomou (2004) explored the offender characteristics of a sample of 21 serial sexual homicide offenders (100% male). The offender’s mean age at the time of the arrest was 32.6 years, 81.0% were previously convicted, 50.0% of the offenders had not finished high school, and 52.4% had a psychiatric diagnosis. Pakhomou (2004) found that 88% of victims were female, the victim’s mean age was 23.3 years, and 70.1% of victims were strangers to the offenders.

Kraemer et al. (2004) examined the demographic characteristics of 157 SH offenders (95% male). He found that the mean age of the offenders was 31.0 years, the mean level of education was 11.5 years, 31% were married at the time of the offence, 16% had been married in the past, and 71% targeted strangers only. The victim’s mean age was 33 years, and the majority of victims (67%) were female.

Salfati and Bateman (2005) reported demographic information on 22 serial SH offenders (100% male). The mean age of the offenders at the time of the start of the series was 32.0 years, 65% of the sample were employed at the time of offence, 48% of the offenders were known to have been engaged in criminal activities, and 70% were single. In this study, Salfati and Bateman found that 52% of victims were female, the victim’s mean age was 28 years, and 31.9% were sex workers.

In a large-scale FBI study (n = 480; FBI, 2008) focusing on SH committed by male offenders in the United States, the authors found that white offenders made up more than half (52.2%) of the sample, and ages ranged from 15 to 45 years at the time of the first homicide. Their average educational level was a high-school diploma or further education (70.2%). The diagnoses of the offenders were mostly personality disorder (42.9%). A large percentage (78.3%) of the offenders had an arrest record at the time of their first homicide. In the FBI study, it was found that the victims were primarily female (75.4%), 31.5% were strangers to the offenders, and in 50.4% the ages ranged from 14 to 29 years.

In the last two decades, some studies were conducted on non-U.S. samples and published in peer-reviewed journals.

Salfati et al. (2015) collected information on 33 SH offenders (64.9% male) in South Africa, from 1953 until 2007. The mean age of the offenders at the beginning of the series was 30.0 years; 89.3% were previously convicted. The majority had a low education (9.7% had no formal education, 48.4% had primary school level of education; 39.3% had high-school education, Grades 8–12). A total of 60% were involved in criminal activity apart from the series of homicide. The majority of victims were female (64.9%), the victims’ mean age was 31.3 years, and 2.3% were sex workers.

Harbort and Mokros (2001), analyzing 61 SH offenders (88.5% male) between 1945 and 1995 in Germany, found that the average age at the time of the first homicide was 27.5 years (SD = 8.2) with ages ranging from 14 to 53 years. Of those offenders, 28.8% had dissocial personality disorder and 17.3% a mixed personality disorder. The average educational level was a basic-level high school (55.2%) or high school (13.8%). A large percentage (79.3%) had a previous criminal record. In this study, victim characteristics were not reported.

In a study conducted on SH committed by 25 SH offenders from 1973 to 2012 in Sweden, Sturup (2018) found that the mean age at first homicide was 29.0 (SD = 8.2). Almost half of the offenders had a personality disorder (46.0%), and more than half had a previous conviction (60%). Slightly more than half of the victims were female (51%), and more than half of the victims (58%) were between the ages of 20 and 64 years. A total of 6% were sex workers.

Recently, Mckinley and Petherick (2021) identified 82 SH offenders (78.9% male), in a time span from 1 January 1820 to 31 January 2020. SH offenders had a mean age of 30.4 years at the time of their first homicide (it should be noted that the mean age was calculated only for those cases, n = 60, with one offender, inclusive of both male and female offenders, and multiple offenders were excluded). The authors stated that most victims were vulnerable because of their isolation, gender, age, drug and alcohol use, or mental health problems, but they did not report descriptive statistics.

Altogether, extant research shows that SH offenders are predominantly male, with an average age of about 30 years at time of the first homicide and tend to have low education (except in a U.S. sample). The majority of them have a personality disorder, and many of them have had previous convictions.

Regarding the classification of types and motives of SH offenders, in her review Miller (2014) concluded that SH is committed by four types of offender: sexual sadistic offenders, who kill to gain an extreme pleasure from torture, control and sexual abuse of the victim; delusional offenders, who are often psychotic and whose homicides are ideologically driven to eliminate people deemed undesirable by them; custodial offenders, who kill by intentional malpractice while working in healthcare; and utilitarian offenders, who kill for financial gain or some other type of material gain and whose motives encompass anger and revenge.

Behavioral crime linking

A number of studies have been conducted on a sample of SH in Italy (Pakkanen et al., 2021; Pakkanen et al., 2012; Salo et al., 2013; Santtila et al., 2008) to evaluate the feasibility of behavioral crime linking, namely the practice of analyzing the crime scene characteristics of two or more offences to determine whether these could share a common offender (for an extensive review on behavioral crime linking, see Woodhams & Bennell, 2015).

These studies indicated that behavioral crime linking was possible. In fact, Santtila et al. (2008) investigated whether the SH offenders were consistent across their series allowing for behavioral linking of homicides committed by the same offender. The authors were able to assign 62.9% of the cases to the correct series in a cross-validation scheme where the chance expectation of the correct assignment was 5.3%. Salo et al. (2013) extended the research by Santtila et al. (2008) by investigating the effectiveness of linking cases of SH using behavioral patterns of offenders, analyzed through Bayesian reasoning. They produced a model that achieved a classification accuracy of 83.6%.

Moreover, Pakkanen et al. (2021) investigated how including both SH and hard-to-solve one-off homicides in the same dataset affected prediction of series membership and found reduced similarity between the one-offs and any series than within the SH. The studies conducted in Italy were limited to SH taking place between 1980 and 2001, including 23 SH offenders.

Taken together, the results of these studies show the consistency of SH offender behavior and the feasibility of conducting behavioral crime linking.

Offender spatial behavior

SH have also been studied from the perspective of the spatial behavior of the offenders. The first attempt to classify SH offenders’ spatial behavior was made by Holmes et al. (1998), who suggested a distinction between geographically stable SH offenders, who live in the same area for some time and kill and dispose of bodies in the same or nearby area, and geographically transient offenders, who travel from one area to the next and dispose bodies in a wider area. Hickey (2015) classified SH offenders into ‘travelers’ who travel from state to state, ‘locals’ who remain stable in their home state, and ‘place specific’ offenders who kill in their own home. Following this classification, Hickey observed that 50% of male offenders fall into the ‘local’ type. Lundrigan and Canter (2001a) found that SH is generally committed farther from offenders’ homes than other types of crimes. However, the home-to-crime distances were shorter for SH rooted in emotion than for SH offenders who chose their locations intentionally (Lundrigan & Canter, 2001b). Snook et al. (2005) found that more than half of SH offenders lived within 10 km of their offence locations and that home-to-crime distances were positively correlated with the offender’s IQ score and negatively correlated with offender age. Synnott et al. (2019) compared SH offenders who chose to dispose of a body and those who did not and found no significant differences between the groups of body‐disposing and non‐body‐disposing SH offenders in the size of the criminal range and in home-to-crime distances. When Synnott et al. compared SH offenders with a sexual motive versus those committing robbery (or ‘acquisitive’), they found no significant differences in the overall median distances between home and crime locations. The only difference in spatial behavior between the two groups of offenders was that the acquisitive offenders were found to be slightly more consistent in their criminal range.

In light of these studies, the study of the offender spatial behavior represents a useful dimension to be considered during the investigation of SHs.

Other studies: time interval between homicides

While the spatial dimension has been studied quite extensively, the temporal dimension has received limited attention. Originally, the time interval between homicides in a series was called the cooling-off period (Burgess, 2006). This term alludes to a process in which the emotions of SH offenders cool down following a killing, as if they entered a refractory period. At the end of this period, which varies from person to person, the fantasies and the emotions that led to the homicide are thought to appear again. The term is now considered obsolete or, as reported in the third edition of the Crime Classification Manual (Douglas et al., 2013, p. 16), it is ‘considered a historical term’. In the current FBI definition of SH offender (Morton & Hilts, 2005) the term ‘separate events’ is used instead. The temporal interval between homicides (TIBH) in SH has received attention by researchers only in recent years.

To our knowledge, considering peer-reviewed studies only, the first study to focus on TIBH was conducted by Lange (1999), who used a cusp catastrophe model to fit to the time series of homicide dates of 11 SH offenders. His analysis focused only on the TIBH, and he excluded the offence characteristics from the analysis. Lange used a polynomial map to express the (n + 1)^th inter-homicide interval through the n^th and (n − 1)^th intervals. After that, by regression analysis he found the coefficients of polynomial that will maximize the correlation coefficient between the actual time series of homicide intervals and time series obtained using the polynomial map. In this way, Lange found evidence for the existence of a cyclic underlying process in the time series and that it was possible to distinguish pairs of SH offenders relying on properties of their time series.

Simkin and Roychowdhury (2014) studied a single case of an SH offender who killed 53 people and found that the distribution of the intervals between homicides followed a power law with the exponent of 1.4. Referring to a study by Osorio et al. (2009) reporting a similar power-law distribution of the intervals between epileptic seizures (with the exponent 1.5), the authors speculated that the condition causing an SH offender to kill arises from the periodical simultaneous firing of large number of neurons in the brain, similar to epileptic seizures. Albeit intriguing, the hypothesis that exceeding a certain threshold of neuronal excitation may cause homicidal behavior lacks theoretical foundation and the empirical support.

Osborne and Salfati (2015) studied 16 homicide series including 90 time intervals from 121 homicides committed by male SH offenders with an offender mean age of 26.63 years (SD = 8.9). They found that the mean TIBH was 186.12 days (SD = 404.4) whereas the median TIBH length across the full sample was calculated as 34.5 days. The shortest series lasted 3 days and the longest 6212 (about 17 years). Moreover, they reported that SH offenders (n = 9) with high levels of social involvement during their series (determined from marital status, occupation, drug/alcohol abuse, gang involvement, criminal history, military history and mental health history) had a higher median TIBH (50 days) than SH offenders (n = 6) without a social involvement (15 days). Those with child victims (n = 2) had a higher median TIBH (367.5 days) than those who killed sex workers (n = 4; 113.0 days) and the elderly (n = 1; 37.0 days).

Edelstein (2020), in a sample of 53 SH offenders retrieved from the Encyclopedia of Serial Killers (Newton, 2006), found that the period (in months) between the first and second homicides (M = 23.32, SD = 40.79) was longer than the period between the second and third homicides (M = 17.15, SD = 36.98); similarly, the period between the second and third homicides was longer than the period between the third and fourth homicides (M = 8.03, SD = 21.30).

Yaksic et al. (2021) used 2837 TIBH for 1012 SH between the years 1901 and 2014 retrieved from the Consolidated Serial Homicide Offender Database (Aamodt, Fox, et al., 2020). They repeated the same statistical analysis as that conducted by Simkin and Roychowdhury (2014) and found that the probability distribution of time intervals between homicides is a smooth, monotonously decreasing function of interval length following a power law in the region of 10–10,000 days.

Sutton and Keatley (2021), analyzing the cooling-off period of two well-known American SH offenders, hypothesized that the taking trophies as well as photographing victims could satiate the offenders’ appetites and appease the urge to kill (it should be noted that they used the cooling-off period term because it is more in keeping with their hypotheses).

In light of these studies, the temporal dimension in the analysis of SHs is a promising field that needs further investigation, including integrating information about modus operandi and spatial behavior.

Aims of the present study

Although several studies have been conducted on SH offenders in Italy, no peer‐reviewed study to date has analyzed the entire population of the SH offenders in the time span from 1848 to 2019. The present study aims to do so by describing the offenders, victims and offence characteristics, thus contributing to improve the knowledge of the phenomenon in Europe and to provide the base rate of several characteristics of SH offenders that can be used as best guess to estimate the characteristics of the unknown authors in SH investigation.

Method

Case selection

We searched for homicides where the same offender(s) committed (upheld judicially) unlawfully at least two homicides, alone or with accomplice(s), with a time interval between the events of at least of 24 hours, without ever being caught by the police during the series of crimes, each of which having to be intentional. Homicides committed in connection with ideological or political terrorism, organized crime or contract killing were excluded. In order to identify cases of SH, Italian criminological literature (true crime books included) and the archives of the most important daily Italian newspapers were consulted using keywords related to SH. In all the cases included in the present research, the offender had been definitively convicted in a court of law, except two offenders who committed suicide before being apprehended. The data set was created anew with all behaviors being coded again and is in this sense not an extension of the data set used in previous studies conducted on SH in Italy (Pakkanen et al., 2021; Pakkanen et al., 2015; Salo et al., 2013; Santtila et al., 2008).

Measures and data coding

The variables used in the present study were for the most part developed based on research by Santtila et al. (2008), which, in turn, were based on research by Salfati (1998). The cases were coded using 14 variables (Table 1) pertaining to the offender, three variables pertaining to the victim (Table 1), 34 variables pertaining to the offence characteristics (Table 2) and seven variables describing other information of the case (season, date and time of the homicide, the number of days between homicides, the date of the arrest of the offender or identification of the offender, and whether the victim and the offender knew each other before the homicide). Cases with missing data on the selected variables were removed from the analyses of only the variable under consideration.

Table 1.

Offender and victim characteristics.

Offender and victim characteristics	M	SD	%	n
Offender characteristics
Age at the first homicide (years)	33.2	11.3
Age during series (years)	35.1	11.3
Male			93.2	55 (0)
Italian			96.6	57 (0)
Alcoholic			17.4	8 (13)
Drug-dependent			10.4	5 (11)
Personality disorder			72.4	42 (1)
Diagnosis of psychosis or a severe personality disorder that caused a total absence of criminal responsibility or a greatly diminished criminal responsibility			27.1	16 (1)
Education up to secondary school			67.3	35 (7)
Stable involvement in a job			68.4	39 (2)
Do not live with a partner during series			51.0	26 (8)
Committed the entire series of homicides after a previous conviction			38.6	22 (2)
Committed the entire series after homicide conviction			8.5	5 (0)
Committed the entire series of homicides while in partial freedom or fugitive			10.2	6 (0)
Victim characteristics
Age (years)	44.0	21.8 (34)
Male			50.4	123 (0)
Sex worker			21.7	53a (0)

Note: N = 59. Numbers in parentheses are missing data.

^a52 female, 1 male.

Table 2.

Offence characteristics and the autocorrelation between them.

Crime scene actions	%	n a	Lag 1autocorrelation
Victim
Male	49.3	106 (0)	.536**
More than 1 victim	22.8	49 (0)	.715**
Victim age	n.a.	n.a.	.663**
SH offender and victim knew each other	38.6	83 (25)	.588**
Sex worker (52 female, 1 male)	24.9	53 (2)	.703**
Homicide place
Homicide at victim’s home	30.5	64 (5)	.710**
Point of the fatal encounter is the crime scene	72.5	153 (4)	.818**
Body disposal
Body found	94.0	202 (0)	.346**
Body found at CS	74.9	158 (4)	.843**
Body covered or hidden	16.6	35 (4)	.695**
Body buried	8.1	17 (4)	1
Body cut into pieces	9.5	20 (4)	.879**
Body burned or partially burned	7.6	16 (4)	.514**
Body or remains thrown into water	5.2	11 (5)	.344**
Body found in the open	55.3	115 (7)	.684**
Body found in countryside or periphery	58.0	120 (8)	.741**
Body found in city	31.1	65 (6)	.694**
Weapon used
Any strangulation	18.1	39 (0)	.556**
Strangulation	13.0	27 (7)	.601**
Choked/strangled	2.9	6 (7)	.863**
Sharp weapon	26.2	55 (5)	.703**
Firearm	42.7	90 (4)	.766**
Other weapon	15.4	32 (7)	.560**
Type of injuries
At least one hit or kick	13.5	17 (89)	.549**
Wounds in only one body part	50.5	93 (31)	.723**
Only damage to throat	10.8	22 (12)	.881**
Multiple wounds in multiple body parts	48.9	90 (31)	.693**
Overkill	23.0	45 (19)	.807**
Restraints
Victim or body tied up	6.3	12 (23)	.380**
Victim tied to something	1.0	2 (23)	–.007
Sexual elements
Genital or breasts exposed	21.2	41 (22)	.514**
Any sexual behavior	28.8	62 (0)	.501**
Sexual relation with victim	17.0	31 (33)	.555**
Rape or rape attempted	6.9	12 (42)	.435**
Post mortem sexual activity	7.1	14 (17)	.494**
Sexual activity evident at the time of the criminal investigation	38.7	24 (153)	.919**
Sexual activity established after arrest	33.9	21 (153)	.931**
Ritualistic elements	7.9	17 (1)	.610**
Messages	9.8	21 (1)	.724**

Note: N = 215 (custodial offenders were not included). n.a. = not applicable. SH = serial homicide; CS = crime scene.

^aNumbers in parentheses are missing data.

**p < .01.

To code offender and victim as well as offence characteristics we consulted daily Italian newspapers and criminological literature (true crime books included). We assessed the reliability of the information for cases that occurred after 1970 by cross-checking court transcripts, police reports and psychiatric examinations, as well as through direct inquiries to the two main police agencies in Italy (Polizia di Stato and Arma dei Carabinieri), public prosecutors and lawyers who were involved in the cases.

For the homicide cases occurring before 1970, due to difficulty in obtaining the trial transcripts, psychiatric examinations and contact information of the law enforcement or judicial officials who were involved in the cases, we coded a variable only if all the available sources indicated the same value for that variable. Otherwise, a missing value was assigned.

As a general rule, we coded variables dichotomously as either present (1) or absent (0). Additionally, we included other general information, such as the age of the SH offender and the dates and time of the homicide. To ensure anonymity, identity information of the offenders and the victims were excluded from the data file.

The data coding of the entire data set was conducted by the first author, who had prior experience with the coding format. To calculate the inter-rater reliability of the coding scheme, we randomly chose one homicide for 10 series, assigning these to a junior researcher who was instructed to code the cases with the coding scheme. Finally, we calculated the inter-rater reliability (IRR) between the coding of the first author and that of the junior researcher. The IRR of the coding scheme was good, with a mean of .776.

We coded offence characteristics for all SH except those committed in care settings, hospitals or healthcare. These we connoted with the term ‘custodial homicide’ (23 homicides) and ‘custodial offenders’ (5 offenders). We analyzed the offence characteristics, the consistency of the offence characteristics over the series, the time-lagged correlation between offence characteristics in pairs of homicides and the consistency of the SH offenders’ behavior just for non-custodial SH offenders.

In the coding of the SH, we considered a sexual element each time that the genital or breasts of the victims were found exposed or when a sexual relation between the SH offender had happened before the homicide or when a rape, an attempted rape or postmortem sexual activity had taken place. We added a variable to assess the presence of written messages from the SH offender (for example, writings found on the bodies of the victims, messages from the SH offender to the police, notes left by the SH offender on a victim’s grave, messages claiming the crime). We marked as a ritualistic behavior each time a crime scene presented unnecessary action for successfully accomplishing the homicide.

We investigated the consistency of the offence characteristics over the series. To do this, we computed all the time-lagged offence characteristics for all series (excluding custodial series), between homicide n committed at time t₀ and homicide n + 1 committed at time t₁. Thus, a higher correlation pertaining to an offence characteristic action indicated that occurrence of the action during a certain homicide (n) was associated with a higher probability of the same action occurring in the subsequent homicide in the series (n + 1).

Suppose we want to calculate the consistency in the use of a firearm as a method of killing. To do that, for all homicide cases (custodial cases excluded), we coded the firearm variable as 1 if a firearm was used in a homicide and as 0 if not.

The correlation matrix consists of m rows and 2 columns, where m represents the number of all pairs of homicides (custodial cases excluded). Cell 1, 1 contains the value of the firearm variable in the first homicide of the first series, and Cell 1, 2 the value of the same variable in the second homicide of the first series. The last row of the matrix (Cell m, 1) contains the value of the firearm variable in the penultimate homicide of the 51 series, and Cell m, 2 the value of the same variable in the last homicide of the last series. We then computed a correlation between Column 1 and Column 2. In this way, a higher correlation pertaining to the firearm variable indicated that occurrence of the use of the firearm during a certain homicide (n) was associated with a higher probability of the use of the firearm in the subsequent homicide in the series (n + 1).

Figure 1 illustrates the method graphically.

Figure 1.

Correlation of the offence characteristic (firearm was used) across all series (n = 51) for each pair of homicides. Series 1 consisted of 4 homicides, whereas Series 51 consisted of 3 homicides. Offence characteristics were coded dichotomously as either present (1) or absent (0). So, the firearm variable was coded as 1 if a firearm was used in a homicide, and as 0 if not.

Results

Offender and victim characteristics

We identified 59 SH offenders who had committed 56 series of homicides. An SH offender has committed two series of homicides. This offender, after being convicted of the first series of homicides (n = 4), committed the second series, n = 4, during the trial period. Fifty-one were exclusively lone serial offenders. Two were considered as a lone serial offender but committed one of their homicides with a co-offender and the rest their series by themselves. Eight SH offenders always committed their offences together with another offender (i.e. four pairs of SH). It should be noted that the numbers of SH offenders that we found differ from those in the Radford University Serial Murderer Database (Aamodt, Leary, et al., 2020), which identified 95 SH offenders in Italy. It should be noted that Sturup (2018) also highlighted the discrepancy between the number of SH offenders he collected in Sweden between 1973 and 2012 (n = 25) and the number reported by the Radford University Serial Murderer Database (n = 10). The large discrepancy between our SH count and that of the Radford database could be explained by a different inclusion criterion.

The majority of the Italian SH offenders included in the present study (n = 50) operated after 1976. Figure 2 shows the numbers of active SH offenders and the number of their victims for each year from 1977 until 2019. The year with the highest number of active SH offenders (i.e. not yet apprehended) was 1992 (n = 12) whereas the year with the greatest number of victims was 1997 (n = 25). The 1990s had the highest number of victims of SH offenders (n = 94) in comparison to other decades (1980s, n = 60; 2000s, n = 27; 2010s, n = 20).

Figure 2.

Active serial homicide (SH) offenders and the number of victims for each year from 1977 until 2019 (SH offenders that acted with an accomplice were counted as 1). 1992 was the year with the highest number of active SH offenders (n = 12). 1997 was the year with the highest number of victims.

Figure 3 shows the trends of the rate of intentional homicide victims per 100,000 people between 1983 and 2018 (ISTAT, 2020) and the fraction of the total that is due to SH. Both trends have been declining since the nineties. The peak of the rate was reached on 1991, whereas 1987 was the year with highest fraction of intentional homicide victims being victims of SH.

Figure 3.

Trends in the rate of intentional homicide per 100,000 people (source: ISTAT, 2020) and the share of homicides attributed to serial homicide (SH) offenders of the total amount of intentional homicide victims from 1983 until 2019.

Of the 59 SH offenders, the majority were men (93.2%, n = 55) and Italian (96.6%, n = 57), and more than half had at least a secondary school education (67.3%, n = 35). The age of the offender during the series was 35.1 years (SD = 11.3), naturally varying during a series. The youngest offender committed his first homicide at the age of 15 years, and the oldest offender committed his last homicide at the age of 62 years. The mean age at the time of the commission of the first homicide was 33.2 years (SD = 11.3). Only a minority of the SH offenders had a substance abuse disorder using either alcohol (17.4%, n = 8) or other drugs (10.4%, n = 5). The majority (72.4%, n = 42) of the SH offenders had a personality disorder, whereas 27.1% (n = 16) had a diagnosis of psychosis or a severe personality disorder that caused a total absence of or diminished greatly criminal responsibility. Regarding work, 68.4% (n = 39) were stably involved in a job (50% were self-employed) whereas 32.2% (n = 19) were not stably involved in a job during the series of homicides (unemployed, retired or students; two offenders, 3.4%, changed their work status during the series). About half of the SH offenders (51.0%, n = 26) did not live with a partner while committing the series of homicides. Many offenders had entered in the criminal justice system prior to all or part of their SH: several of them committed the entire series of homicides after a previous conviction (38.6%, 22), a few (8.5%, n = 5) committed the entire series of homicides after a conviction for homicide, and others (10.2%, n = 6) committed the entire series of homicides while in partial freedom or as fugitives. We also found that a few characteristics of the loner serial offenders (n = 46) changed during the progress of their series. For example, two SH offenders changed their work status, and three of them changed their relationship status.

The SH offenders murdered in total 244 victims (221 victims were murdered by non-custodial offenders). About half (50.4%, n = 123) of the victims were male. The mean age of the victims was 44.0 years (SD = 21.8), ranging from the youngest being 4 years-old to the oldest being 91 years-old. For 34 of the victims (13.9%), the age remained unknown. Of victims, 21.7% (n = 53) consisted of sex workers (52 female and 1 male). Table 1 presents a list of offender and victim characteristics.

Offence characteristics

We found that in almost one out four cases there was more than one victim (22.8%, n = 49). The bodies of the victims were found in most cases (93.9%, n = 202), frequently on the crime scene (74.9%, n = 158), in the open (55.3%, n = 115) and in countryside or in the periphery of cities (58.0%, n = 120). In 98.2% of these cases we identified the weapon used: firearm 42.7% (n = 90); sharp weapon 26.2% (n = 55); strangulation by an object/suffocation/strangulation by hands, 18.1% (n = 39); other weapon 15.4% (n = 32). Victims were injured only in one body part 50.5% (n = 93) of the times, and an overkill was observed in 23.0% (n = 45). Sexual elements were found in 28.8% (n = 62).

Table 2 lists the offence characteristics and the time-lagged correlations between them in pairs of homicides.

We next explored the consistency of the non-custodial serial offenders’ behaviors. We did this by counting how many non-custodial offenders enacted the same actions and how many varied in their actions during the series. As can be seen in Table 3, there is high variability in the SH offenders’ behavior, including the type of victim (32.0% male victim, 18.0% more than one victim, 18% sex worker as a victim), the body disposal (30% body was found in the open), the weapon used (24.0% any type of strangulation), the type of injuries (28.9% wounds in only one body part) and the presence of sexual elements (32% any kind of sexual behavior).

Table 3.

Consistency of the behavior of SH offenders.

	Always		Never		Variable
Offence characteristics	%	n	%	n	%	n
Victim
Male	24.0	12 (0)	44.0	22 (0)	32.0	16 (0)
More than 1 victim	8.0	4 (0)	74.0	37 (0)	18.0	9 (0)
Sex worker	20.0	10 (0)	62.0	31 (0)	18	9 (0)
SH offender and victim knew each other before the day of the homicide	44.7	21 (3)	31.9	15 (3)	23.4	11 (3)
Homicide place
Homicide at victim’s home	14.0	7 (0)	56.0	28 (0)	30.0	15 (0)
Point of the fatal encounter is the crime scene	59.2	29 (1)	28.6	14 (1)	12.2	6 (1)
Body disposal
Body found	84.0	42 (0)	2.0	1 (0)	14.0	7 (0)
Body found at crime scene	60.0	30 (0)	26.0	13 (0)	14.0	7 (0)
Body covered or hidden	16.0	8 (0)	66.0	33 (0)	18.0	9 (0)
Body buried	10.0	5 (0)	90.0	45 (0)	0	0
Body cut into pieces	8.0	4 (0)	86.0	43 (0)	6.0	3 (0)
Body burned or partially burned	4.0	2 (0)	84.0	42 (0)	12.0	6 (0)
Body or remains thrown into water	0.0	0 (0)	86.0	43 (0)	14.0	7 (0)
Body found in the open	40.0	20 (0)	30.0	15 (0)	30.0	15 (0)
Body found in countryside or periphery	44.0	22 (0)	30.0	15 (0)	26.0	13 (0)
Body found in the open	40.0	20 (0)	30.0	15 (0)	30.0	15 (0)
Body found in countryside or periphery	44.0	22 (0)	30.0	15 (0)	26.0	13 (0)
Body found in city	18.0	9 (0)	58.0	29 (0)	24.0	12 (0)
Weapon used
Any strangulation	12.0	6 (0)	64.0	32 (0)	24.0	12 (0)
Strangulation	8.0	4 (0)	76.0	38 (0)	16.0	8 (0)
Choked or strangled	4.0	2 (0)	94.0	47 (0)	2.0	1 (0)
Sharp weapon	18.0	9 (0)	64.0	32 (0)	18.0	9 (0)
Firearm	26.0	13 (0)	56.0	28 (0)	18.0	9 (0)
Other weapon	10.0	5 (0)	72.0	36 (0)	18.0	9 (0)
Type of injuries
At least one hit or kick	20.0	6 (0)	66.7	20 (0)	13.3	4 (0)
Wounds in only one body part	36.4	16 (6)	38.6	17 (6)	25.0	11 (6)
Only damage to throat	4.1	2 (1)	87.8	43 (1)	8.2	4 (1)
Multiple wounds in multiple body parts	36.4	16 (6)	36.4	16 (6)	27.3	12 (6)
Overkill	18.4	9 (1)	71.4	35 (1)	10.2	5 (1)
Restraints
Victim or body tied up	4.2	2 (2)	85.4	41 (2)	10.4	5 (2)
Victim tied to something	2.1	1 (2)	95.8	46 (2)	2.1	1 (2)
Sexual elements
Genital or breasts exposed	16.0	8 (0)	58.0	29 (0)	26.0	13 (0)
Any sexual behavior	20.0	10 (0)	48.0	24 (0)	32.0	16 (0)
Sexual relation with victim	7.5	3 (10)	70.0	28 (10)	22.5	9 (10)
Rape or rape attempted	7.7	3 (11)	82.1	32 (11)	10.3	4 (11)
Post mortem sexual activity	4.3	2 (4)	82.6	38 (4)	13.0	6 (4)
Sexual activity evident at the time of the criminal investigation	34.6	9 (24)	53.8	14 (24)	11.5	3 (24)
Sexual activity established after arrest	19.2	5 (24)	73.1	19 (2)	7.7	2 (24)
Ritualistic elements	8.0	4 (0)	82.0	41 (0)	10.2	5 (0)
Messages	6.0	3 (0)	82.0	41 (0)	12.0	6 (0)

Note: Custodial offenders were not included. Numbers in parentheses are missing data. Serial offenders who operate with an accomplice were counted as 1. SH = serial homicide.

Temporal perspective

The overall mean number of months between first offence and apprehension of the SH offender was 39.7 months (SD = 49.9, min = 0.1, max = 205.0, median = 25.1) whereas the overall mean number of months between first offence and last offence was 28.3 months (SD = 40.9, min = 0.0, max = 188.4, median = 10.7). We explored the TIBH across the full sample. The mean TIBH was 10.3 months (SD = 19.5, min = 0.0, max = 149.6). The distribution was positively skewed (skewness statistic = 3.9), meaning that is has a long right tail so the median is a more appropriate measure of central tendency. Median overall TIBH length was 2.8 months.

Comparison of our results with other studies conducted on samples of European serial homicide offenders

When we compared our SH offenders’ descriptive statistics to those in the other two studies conducted in Europe (Harbort & Mokros, 2001; Sturup, 2018; Table 4), we found a few similar results, particularly in terms of gender, age at the first homicide, work status and education. It should be noted that descriptive characteristics were not available for all offenders. The comparison of the characteristics of the offenders in our study with those in other European studies must take into account the different SH definitions adopted (Table 4). For instance, Sturup (2018) included contract offenders in the Swedish sample, and Harbort and Mokros (2001) excluded SH committed in conjunction with robbery or rape, as well as SHs of patients from their sample. Moreover, they excluded homicide cases where the offender suffered from severe hallucinatory schizophrenic psychosis: this could partly explain the high percentage of offenders with a personality disorder in the German sample.

Table 4.

Offender, victim and offence characteristics in three European countries.

		Study/country/time span
		Sturup (2018)a	Harbort and Mokros (2001)b	Current studyc
		Sweden	Germany	Italy
		1973–2012	1945–1999	1848 − 2019
SH offenders (N)		25	61	59
Victims (N)		63	n.a.	244
Share of homicides attributed to serial offenders (since 1983–2012)		1.6 % (63, 3933)d	n.a.	0.6 % (175, 27,771)d
SH offender male		84% (21)	88.5% (61)	96.6% (57)
SH offender: age in years at the first homicide [M (SD)]		29.0 (8.2)	27.5 (8.2)	33.2 (11.3)
Substance abuse		48% (12)	n.a.	Alcohol 13.6% (8) Drug 8.5% (5)
IQ		<70 4% (1) 70–79 29% (7) 80–119 42% (10) 120–129 17 % (4) >130 8% (2)	70–79.9, 7.1% 80–89.9, 10.7 % 90– 99.9, 25% 100–109.9, 42.9% 110–119.9, 8.9 % > 120 5.4%	n.a.
Personality disorder		46% (11)	88.5% (46)	71.2% (42)
Schooling		n.a.	Institution for pupils with learning disability 18.9 % Grammar school 12.1 % Basic level high school 55.2 % High school 13.8 %	Elementary or secondary school 59% (39)
Unemployed		n.a.	39.3%	32.2% (19)
Criminal record: previous convictions		Previous convictions 60 % Previous convictions for violence offence 44 %	79.3 %	45.8 % (27)
Victims
Gender Male		49% (27)	n.a.	50.1% (123)
Age (years)	0–15	17% (9)	n.a.	4.6% (11)
	16–19	8% (4)		3.7% (9)
	20–29	11% (6)		18.0% (44)
	30–44	32% (17)		22.1% (54)
	45–64	15% (8)		18.8% (46)
	> 65	17% (9)		30.7% (75)
Stranger victim		43% (23)	n.a.	45.6% (112)
Sex Worker		6% (3)	n.a.	21.7% (53)
Temporal perspective
Weeks between first and last offence, mean [M (SD)]		98.4 (136.7)	n.a.	122.8 (177.9)
Weeks between first offence and apprehension, mean [M (SD)]		163.9 (249.1)	n.a.	172.5 (216.9)
Modus Operandi
Body found outside		38% (20)	n.a.	55.3% (115)
Weapon used		Firearm 13% (7) Asphyxia 45% (24) Knife/axe 19% (10) Other 11% (6)	Available data only for subset of serial sexual homicide	Firearm 40.7% (90) Sharp weapon 24.9% (55) Strangulation by an object/Soffocation/Strangulation by hands, 17.65%, (39) Other 15.3% (33)
Sexual elements/motive		32%	41%	28.8%

If not otherwise specified, numbers in parentheses are frequencies. SH: serial homicide; n.a.: not available.

The first number is the total number of victims of serial offenders, the second represents the number of individuals who died as a result of homicide.

^a Sturup (2018) – Definition: Individuals who committed two or more homicides over a period exceeding 24 hours. Contract murderers were included.

^b Harbort & Mokros (2001) – Definition: The fully or partially culpable perpetrator commits alone or with accomplice(s) at least 3 completed murders, each premeditated and characterized through a new, hostile intent. Murders in conjunction with robbery or rape were excluded. Perpetrators who suffered with severe hallucinatory schizophrenic psychosis were excluded. Serial murders of patients were not included.

^c Current study – Definition: Homicide cases where the same serial homicide offender(s) committed (upheld judicially) unlawfully, alone or with accomplice(s), at least two homicides in separate events, with a time interval between homicides of at least 24 hours, each homicide intentional. Homicides committed in connection with ideological or political terrorism, organized crime or contract killing were not included. Serial homicide of patients was included.

^d The first number is the total amount of the victims by serial murderers, the second number represents the number of the individuals died as the result of homicide.

Discussion

To our knowledge, this is the first study attempting to describe the offender and victim characteristics of SH over such a long period of time in Italy and in a European country more broadly.

To study this phenomenon, one of the preliminary issues is to choose a definition of SH. We considered the number of the victims, the time interval between homicides and the primary motive of the homicide (see Adjorlolo & Chan, 2014, for a detailed discussion of the controversies regarding SH definitions). In our study, we collected cases where the same offender(s) committed (upheld judicially) unlawfully alone or with accomplice(s) at least two homicides with a time interval between the events of at least of 24 hours, without ever being caught by the police during the series of crimes, each of which had to be intentional. Homicides committed in connection with ideological or political terrorism, organized crime or contract killing were excluded.

The year 1997 had the highest number of SH victims, and 1992 had the highest number of active serial offenders in Italy. To our knowledge, the trend of the phenomenon is unknown in Europe, whereas in the United States the decline of the SH began in the 1980s (Yaksic et al., 2019). In Italy, the rate of intentional homicides per 100,000 people has been in decline since 1991 when it reached 3.38. In the year 2018, Italy had the fourth smallest incidence of intentional homicides in Europe (0.57 per 100,000 people; EUROSTAT, 2020). The steady reduction in the number of homicides for terrorism and organized and common crime is the factor that has contributed to the decrease in the number of homicides; however, in our dataset, these two types of homicide were not included.

Even if it has been observed that violence has been in decline over millennia and that the present is probably the most peaceful time in the history of the human species (Pinker, 2011), there is no reason to think that the psychological and psychopathological conditions that led SH offenders to kill have generally changed in a ‘favorable’ sense in the last decades. Therefore, the decrease of SH in Italy may not be explained by a reduction in the number of potential SH offenders in recent years but, rather, by an improvement in the investigative techniques that allow the police to detect potential SH offenders already after their first offence. Therefore, it could be speculated that the decline of SH could be due to developments in law enforcement techniques, technology and the increase in personal safety measures over time; however, empirical research is needed to reach this conclusion (Yaksic, 2019).

We found that the values of our SH offenders’ descriptive statistics were similar to those from the other two studies conducted in Europe (Harbort & Mokros, 2001; Sturup, 2018; Table 4), we found a few similar results, particularly in terms of gender, age at the first homicide, work status and education. It should be noted that descriptive characteristics were not available for all offenders. The comparison of the characteristics of the offenders in our study with those from other European studies must take into account the different SH definitions adopted (Table 4). For instance, Sturup (2018) included contract offenders in the Swedish sample, and Harbort and Mokros (2001) excluded SH committed in conjunction with robbery or rape, as well as SH of patients from their sample. Moreover, they excluded homicide cases where the perpetrator suffered severe hallucinatory schizophrenic psychosis: this could partly explain the high percentage of offenders with a personality disorder in the German sample.

When we analyzed the crime scene of all homicides, excluding those committed in care settings, we found numerous significant correlations between offence characteristics for one homicide and those for the subsequent one. The correlations concerned victim characteristics and modus operandi (homicide place, body disposal, weapon used, type of injuries, and sexual and ritualistic elements).

These correlations suggest a consistency of the offence characteristics over the series. However, this current study does not allow us to reach a conclusion on the consistency of the SH offenders’ behaviors in committing their series of crimes because we computed all the time-lagged offence characteristics for all series, between homicide n committed at time t₀ and homicide n + 1 committed at time t₁. In other words, we assessed the consistency of all offence characteristics for each pair of homicides across the whole series of homicides and not the consistency of the single SH offenders’ behaviors.

Regarding the issue of the consistency of the SH offenders’ behavior, there have already been studies supporting the consistency of SH (Lundigran & Canter, 2001b; Salfati & Bateman, 2005), and the consistency of the behaviors of Italian SH offenders was observed in two previous studies conducted on a subset of the population of this current study (Salo et al., 2013; Santtila et al., 2008). Previously, it was assumed that less situationally dependent behaviors would be more consistent (Bennell & Canter, 2002); however, Woodhams and Labuschagne (2012) found that the making linking predictions using statistical algorithms, they got better performance when predicting using all offence characteristics of the modus operandi and not one particular domain.

The consistency of the offender’s behaviors can be explained in the light of person–situation interaction patterns by several authors (Greene, 1989; Meyer, 1990; Mischel & Shoda, 1995). According to these models, people interpret the events they encounter through a system of mental representations (objectives, strategies, beliefs, plans), which, in turn, result in behaviors. An individual’s behavior is therefore the result of interpreting the situation and remembering how he or she acted and faced a similar situation previously.

The more frequently a behavioral strategy is activated, the more likely it is to be activated in the future (Green, 1989). Therefore, according to these models, behaviors are a combination of a person’s current goal, their past learning and the situational characteristics.

It is important to note that the similarity of situations refers also, and above all, to the psychological similarity perceived by the person (Shoda, 1999). In other words, even if a person is faced with a situation objectively similar to a previously encountered one, it does not mean that he will act in exactly the same way precisely because it depends on the subjective interpretation with which the person reads the event by his unique system of mental representations. This factor, the victim’s reactions, the offender’s expertise learned from previous homicide, the offender’s potential contact with the police and legal processes (Davies, 1992) and other unexpected events that may occur during the commission of the crime cause a variability in the offender’s behavior (about the psychology of linking crime see Woodhams et al., 2007).

In our study, when we counted how many SH offenders have always acted with the same behavior in all the crimes of their series, we observed a variability in their behavior. For example, the observation that 32% of the SH offenders sometimes had sexual elements in their crimes and sometimes not can be explained in light of the aforementioned factors and of the explanatory models of the variability and consistency of human behavior. Therefore, our results show that the presence or absence of a sexual element at the crime scene should not lead investigators to conclude, only on the basis of this crime feature, that a homicide belongs or does not belong to a series of homicides. Moreover, our results show the criticality of the method of classification of SH offender based on the presence of sexual elements at the crime scene. In other words, since an SH offender can show sexual elements in one crime and not show them in another, then they could jump from one category (sexual homicide) to another (non-sexual homicide) and vice versa depending on whether a homicide of his series shows sexual elements or not. The variability in the behavior of SH offenders was also observed by Harbort and Mokros (2001) who found that the majority of the offenders changed the modus operandi from one homicide to the next.

Even the ritualistic elements and postmortem sexual activity sometimes can appear in a series but may not always be present. Offence characteristics that show postmortem activity can be considered elements of a signature – that is, an expression of violence-oriented fantasies that are of great importance for the offender (Douglas et al. 2006; Hazelwood et al., 1992; Keppel, 1995, 2000). Therefore, one might expect that they will always be present at the crime scene. However, in our population, the number of those who always exhibited ritual behaviors (4) was equal to those who exhibited them only sometimes. Even Schlesinger et al. (2010), in their sample of 38 U.S. SH offenders, found that the offence characteristics of the offenders are fairly complex and varied and that they rarely engaged in exactly the same behavior at every homicide. The emergence of these behaviors only sometimes, the appearing and disappearing like a karst river, during the series of crimes could be due to contingent circumstances that prevent the offender from carrying out their fantasies. For example, the offender did not have enough time to carry out his behavior or the victim’s reactions may have forced the offender to control the victim in a way he otherwise would not have done. The offender may also have altered his behavior due to changes in the aims of the offence during the series of crimes or as a result of learning.

It should be noted that not only can the modus operandi change, but also some of the characteristics of the offender. And this must be taken into consideration in the prediction of the characteristics of an unknown offender (Alison et al., 2002). Of course, the age of the offender is a characteristic that changes with certainty and in a known direction, if the SH offender continues to be active. So, if the date of the first homicide of the series is known then, after having estimated the offender’s age for that crime, it will be possible to estimate the age in the next homicide committed by the same offender. Nevertheless, other dichotomous characteristics that do not follow the laws of physics do not have an obligatory direction. We found, for example, that two offenders changed their work status, and three changed their relationship status.

Our results show that not all SH offenders are uniquely sexually motivated and that, as we said above, some offenders can sometimes exhibit an objectively detectable sexual element in a crime in the series and sometimes not. This has already been observed about 20 years ago. Hickey (2015) warned that to consider SH as exclusively sexual in nature is a misrepresentation, and the FBI (2008) stated clearly that SH is not synonymous with sexual homicide. Our findings went in that direction. We found sexual elements in about one out of three of the series. A similar finding was observed by Sturup (2018) in the Swedish study, whereas in the German study a higher percentage was observed (still below 50%). We found that about half of victims were male, and these findings are similar to Sturup’s results. The percentage of sex worker victims (about 1 out of 5) was greater than that in the Swedish study. Sturup speculated that the low percentage of sex workers as victims that he found could be due to the fact that prostitution in Sweden has more or less disappeared from the streets and has moved into contexts controlled by organized crime, which makes it more unlikely that sex workers will become victims of a lone SH offender wandering around the city at night. Anyway, even in Italy sex workers are controlled by organized crime but the streets are a place where sex workers are still present.

Peer-reviewed studies analyzing the temporal dimension in SH are still few. When we compared our results to those of the only European empirical research that reported this information (Sturup, 2018), we found a difference in the time length between first and last offences (greater in our study) and a difference in mean length between first offence and apprehension (greater in our study).

The present study lays the foundation to conduct further studies in order to improve the knowledge to apply to SH investigations particularly on behavior crime linking, which has been shown to be viable (i.e. utilizing information concerning offence characteristics in order to link crimes committed by the same offenders). Moreover, the addition of geo-referenced data (such as the coordinates of the offender’s home base, the body dump location or the location of the point of encounter between an offender and a victim) would make it possible to evaluate whether combining the offender’s spatial behavior and his offence characteristics could improve offender profiling and the behavioral crime linking, and also to verify the feasibility of the geographic profiling (i.e. the prediction of the offender’s home base location).

Limitations

The study has a number of limitations. First, our ambition to describe the entire population of Italian SH offenders collides with the effective possibility of tracing all cases. That originates for two reasons: (a) despite the consultation of two online archives of two of the main national newspapers, the reading of all Italian criminology books (true crime books included) and the personal request addressed to the Arma dei Carabinieri and Polizia di Stato, we expect that some cases, especially the old ones, could have eluded our data collection attempts; (b) a few unsolved homicide cases committed by the same offender may not have been linked by the law enforcement, and therefore a few homicide series, or just unsolved isolated cases, potentially remained out of our data set.

Secondly, only solved homicide series (upheld judicially) were included. As Bennell and Jones (2005) point out, it may be possible that solved cases show higher levels of behavioral stability and distinctiveness than unsolved cases.

A third limitation is pertinent to a potential coding bias that could have affected the principal investigator of this study during the case analysis. However, Pakkanen et al. (2012) tested whether coders’ prior knowledge of which crimes had been committed by the same SH offender affected the similarity that these coders perceived in the behavior of the offender in linked crimes. They found no evidence for the presence of a coding bias. In other words, prior knowledge about which crimes had been committed by the same offender did not distort participants’ perceived (i.e. coded) behavioral similarity in favor of their expectations.

Conclusions

The present study outlined the phenomenon of SH in Italy from 1849 to 2019 in terms of offender and victim characteristics and offence characteristics improving the knowledge of the phenomenon in Europe. The frequencies of the characteristics reported for the SH offenders can be used as best guess to estimate the characteristics of the unknown authors in SH in ongoing investigation in Italy.

Ethical standards

Declaration of conflicts of interest

Angelo Zappalà has declared no conflicts of interest

Shumpei Haginoya has declared no conflicts of interest

Pekka Santtila has declared no conflicts of interest

Ethical approval

This article does not contain any studies with human participants or animals performed by any of the authors.