Delayed prosecutions of child sexual abuse (CSA) is “now a thriving legal industry” stated Madam Justice Southin of the British Columbia Court of Appeal (R. v. R. (J.W.), 2001, at para. 26). The Select Committee on Home Affairs in the U. K stated “[i]n the last 5 years, 34 of the 43 police forces in England and Wales have been involved in investigations into allegations of child abuse in children's homes and other institutions. All of the allegations relate to historical abuse, said to have occurred several years—often decades—ago” (Home Affairs Committee, 2002). Even in the United States, where most jurisdictions have limitation periods on criminal prosecutions, recently there have been discussions of whether old crimes should be prosecuted (e.g., Alpert et al., 1998; Haber & Haber, 1998; Shuman & McCall Smith, 2000). Given that most child sexual abuse victims do not disclose in a timely fashion (for a review see London, Bruck, Ceci, & Shuman, 2005) the question of prosecution of old crimes may have particular relevance to CSA cases. This issue raises complex questions concerning, inter alia, research and the development of public policy. Informed discussions, we believe, must begin with an understanding of the phenomenon. Thus, the purpose of this paper is to inform and expand the exchange of ideas through a detailed description and discussion of delayed prosecutions of CSA (historic child sexual abuse or HCSA).

Despite the fact that the vast majority of historic cases involve claims of continuous memory of the abuse, the literature on delayed prosecutions of CSA focuses on cases involving claims of memory repression. Moreover, much of the research on continuous memory of prior sexual abuse concerns delayed disclosure rather than delayed prosecution. We review the literature on delayed disclosure for two reasons: because disclosure of the abuse is a necessary, albeit not sufficient, condition for delayed prosecution and because understanding delayed disclosure provides a starting point for studying delayed prosecution.

Research with community samples of adults in Canada and the US has concluded that delayed reporting of child sexual abuse is surprisingly common. For example, the Report of the Committee on Sexual Offences Against Children and Youths (hereinafter Badgley Report, 1984) reported that 76.2% of women and 88.9% of men who reported having been abused as children responded “no” when asked if they had ever reported these incidents; Lamb and Edgar-Smith (1994) estimated that 64% of CSA victims fail to report their abuse during childhood (for similar estimates see Roesler, 1994; Roesler & Wind, 1994); and Finkelhor, Hotaling, Lewis, and Smith (1990) found that 33% of women and 42% of men who reported having been sexually abused as children had not disclosed the abuse before the survey. More recently, Arata (1998) found that 69% of CSA victims did not disclose the abuse during childhood, while Smith et al. (2000) reported that 47% of the adults they studied who reported having been abused as children had not reported the abuse for more than 5 years and a further 27% reported that they disclosed their abuse for the first time on the survey.

Scholars have presented both quantitative and qualitative research to explain delayed disclosure of CSA. Quantitative predictors of delay that have been studied are intrusiveness of the abuse, relationship with perpetrator, age of child at the time of the abuse, frequency/duration, and gender of the child. With the possible exception of gender, all of these variables are unstable as predictors of delay. Arata (1998) and Badgley (1984) found that more intrusive abuse predicted delay (Smith et al., 2000) found that more intrusive abuse predicted immediate disclosure, and Lamb and Edgar-Smith (1994) and Roesler (1994) reported that intrusiveness and timing of disclosure were unrelated. Relationship with the perpetrator is similarly unstable as a predictor of delayed disclosure. Arata (1998) and Smith et al. (2000) reported a longer delay if the perpetrator was a family member, and particularly a person in a parent-role while Lamb and Edgar-Smith (1994) reported that the child's relationship with the perpetrator was unrelated to the timing of disclosure. A similarly unstable outcome is found in studies of age as a predictor of delay. Smith et al. (2000) reported that a younger age at the time of the abuse was predictive of a longer delay to disclosure whereas Arata (1998) reported that the child's age was unrelated to delay. In studies of the effect of duration/frequency as predictors of delay, Arata (1998) reported a trend toward longer duration of abuse predicting a longer delay and Smith et al. (2000) found that multiple offences predicted a longer delay. On the other hand, Lamb and Edgar-Smith (1994) found neither a relationship between duration nor frequency and delay. The one factor that is most consistently predictive of delay, however, is gender; boys are less likely to disclose than girls (Badgley, 1984; Finkelhor et al., 1990; Lamb & Edgar-Smith, 1994, but see Goodman-Brown, Edelstein, Goodman, Jones, & Gordon, 2003). In sum, although there is limited consensus on the factors that predict delay, there is consensus that delayed disclosure of CSA is common.

Interviews with adults who report CSA provide several reasons for delayed disclosure. The most common reasons given for failure to disclose immediately were: “too ashamed it happened,” “too personal a matter to tell anyone,” “afraid of the person who did it,” and, for males “wasn't important enough” (Badgley, 1984, p. 189). Roesler and Wind (1994) reported the three most common reasons given by adult survivors for failure to report CSA during childhood were shame, repression, and fear for safety. Paine and Hansen (2002) reported the following reasons for failure to report the abuse (in no particular order): feeling responsible for the abuse, shame and stigma, fear of being blamed and/or negatively judged, hesitance to break the promise to keep the secret, fear of not being believed, and fear due to threats made by the perpetrator.

Many victims of CSA will wait until adulthood to disclose the abuse, and some of those will pursue a legal remedy.Footnote 1 Although the particular reasons for the delay are still being investigated, some are likely characteristics of the offence itself (e.g., gender of the complainant, a dependent relationship between the child and the perpetrator) and as such are enduring. To the extent that these factors help to explain delayed disclosure, the high percentage of CSA victims who delay disclosing the abuse for a very long time is not a transient social anomaly that will abate over time. Thus, absent the formation of new legal barriers to the prosecution of HCSA cases, we expect that courts will continue to adjudicate such cases (see also Connolly & Read, 2003).

Subject to legal constraints (e.g., statutes of limitations, discussed below), complainants can proceed civilly or criminally. It is beyond the scope of this paper to debate the relative merits of proceeding one way or the other; however, two points are important. First, generally, when a case proceeds civilly the complainant must retain and pay counsel whereas when a case precedes criminally the Crown/State funds the prosecution. Second, generally, it only makes sense to proceed civilly if the accused is solvent. Punishment, rather than monetary compensation, is the purpose of criminal law and so insolvency is not a constraint to proceeding. Arguably, then, criminal cases represent a wide range of complainants—those who could and those who could not afford to finance their own legal proceeding as well as cases against solvent and insolvent defendants. The data that we present represent criminal prosecutions and accordingly, we argue, a broad range of complainants.

Statutes of limitations on criminal offences present a significant obstacle to prosecution. Very briefly, statutes of limitations prescribe that the formal legal proceeding must commence within a specific period of time from some triggering event (e.g., the last occurrence of a recurring offence). Notably, statutes of limitations can be tolled (i.e., suspended) in certain circumstances (e.g., until the victim reaches the age of majority). In most American states there is a statute of limitations on criminal offences; however, there are several notable exceptions: Wyoming does not have a criminal statute of limitations; Kentucky, Virginia, and West Virginia do not have statutes of limitations on felonies (child sexual assault is a felony); Alaska, Maryland, North Carolina, South Carolina, and Rhode Island do not have statutes of limitations on most sexual offences involving children; and in Alabama and Maine there is no statute of limitations on serious sexual offences against children under the age of 16. Moreover, in almost all states that have a statute of limitations on sexual offences against children, it is tolled until the child reaches the age of majority (National Center for Victims of Crime, 1998; Shuman & Smith, 2000). In Canada, as in the United Kingdom, Australia, and New Zealand, there is no limitations period on indictable (roughly equivalent to felony) offences, except in very rare circumstances. Although the data that we present in these papers are from Canadian courts, the phenomena and insights the data provide are of international interest.

The data set presented in this paper contains 2064 criminal complaints of HCSA. Each case was coded on the variables listed in Appendix A related to the legal context, the complainant, the complaint, and the accused. Our intent was to describe the cases as comprehensively as possible and so we coded all of the variables that judges tend to discuss in their decisions. Many of the variables related to the complainant, the accused, and the complaint parallel the variables that have been studied in the context of delayed and non-disclosure of CSA. Variables related to the legal context, for instance trial outcome and inclusion of an expert witness, are basic to the trial process and are included as dependent variables in empirical CSA research.

Sorenson and Snow (1991) have argued that the ability to develop effective guidelines and protocols concerning CSA must rest to some degree on understanding the phenomenon. The study of HCSA legal cases would benefit similarly, we believe, from a careful and thorough description of the phenomenon. To that end, we provide a comprehensive description of a large database of HCSA cases with particular attention to the variables that have been included in studies of CSA. To the best of our knowledge, this data set is the only one of its kind, although the methodology is strikingly similar to Groscup, Penrod, Studebaker, Huss, & O'Neil (2002) in a study of the effect of Daubert on expert testimony admission decisions. As a result, we are able to offer unique perspectives on the forensic experience with HCSA prosecutions. We hope that these data and analyses will inspire further research, discussion, and debate in the areas of research and the development of public policy. Accordingly, in the discussion, we introduce a number of related issues that would benefit from such a scholarly exchange of ideas.

METHOD

QuicklawTM was used to locate criminal cases of delayed allegations of child sexual abuse. QuicklawTM is a full-text data base that contains, at least from 1986 forward: all Supreme Court of Canada decisions; decisions from provincial Courts of Appeal; written decisions from the provincial Superior Courts; and written decisions from Provincial Courts that were forwarded to QuicklawTM. Although we do not know with certainty why judges forward decisions to QuicklawTM, it is fair to say that decisions that judges believe should be available to the legal community were forwarded. Only English decisions were included (i.e., exclude all decisions from Quebec and some from New Brunswick). Importantly, this is not an exhaustive set of criminal HCSA cases heard between 1986 and 2002. The data set, however, contains most decisions that are available to the political, legal, and research communities upon which policy, practice, and research funding allocation decisions will be made.

Search Strategy

We have not been able to find a precise legal definition of “historic” other than rather vague references to “years or decades.” Nor have we been able to find a generally accepted definition of “delay” in the psychological literature. Accordingly, our first task was to define the term “historic.” Because our concern was with “stale” legal claims, statutes of limitations offered guidance. Civil statutes of limitations on personal injury actions (excluding sexual assaults against children because public policy dictates special provisions in many jurisdictions) were the most appropriate because it would allow us to research relevant law in Canada and United States. In Canada, in eight of the nine common-law provinces and in all three territories, the limitations period on personal injury actions is 2 years (see CANLII, 2005 for access to each statute of limitations). In 21 of the 51 jurisdictions in the United States (50 states plus the District of Columbia) the limitations period on civil cases involving personal injury is 2 years (Nolo, 2005). To the extent that statutes of limitations exist because the passage of time leads to loss of evidence including memory decay, the appropriate time to start the limitations clock is when the action under investigation ends. Accordingly, in both Canada and the United States, if the offence is one that recurs, the limitations period begins when the action that is under investigation ends. For our purposes, then, a case was “historic” if 2 years or more had elapsed from the end of the offence to trial.

There is substantial variability across jurisdictions and across time in terms of the length of delay between an official complaint and trial. By using delay to trial, we will have overstated the actual delay from the end of the offence to the official complaint by the amount of time it took to get to trial. However, because it was unusual for judges to report either the date of the official complaint or the charge date, trial date was the most reliable data we could obtain. Arguably, trial date is the most relevant lag for some purposes, e.g., the actual time between the event itself and the complainant's report of it in court.

A key-word search strategy was used including the words “child” (and variations) and the following sexual offences: “sexual offence(s),” “sexual assault,” “sexual interference,” “sexual intercourse,” “gross indecency,” “indecent assault,” “incest,” “rape,” “bestiality,” “buggery.” Each case was reviewed to confirm that the complainant was 19-years-old or younger when the alleged offence began and that two or more years had elapsed from the end of the alleged offence to the trial date. There were two waves of data collection. In the first wave, cases released between 1986 and 1998 were collected: there were 874 cases involving 1626 complainants. Intercoder agreement was obtained on cases involving 167 complainants. Two coders were involved for the purposes of obtaining intercoder agreement and one of these persons coded the remaining cases. In the second wave of data collection, cases from the latter part of 1998 to 2002 were gathered: the search located 228 cases involving 438 complainants and engaged two new coders. Intercoder agreement was obtained on a subset of these cases involving 50 complainants. Once acceptable agreement was obtained, the cases were evenly distributed between the two coders.

Intercoder agreement was computed as (agreements/agreements + disagreements)×100 (i.e., percentage agreement). When dates were coded, an agreement was recorded if both coders were within 1 year of each other. On all variables, a disagreement was recorded if one coder recorded information about the detail and the other coder recorded it differently or recorded it as missing data. This latter source of disagreement explains the minor deviations from perfect agreement on variables that seem self-evident (e.g., trier of fact). Intercoder agreement was not computed for derived variables; length of delay was computed as trial date−date the abuse ended, duration was computed as date the abuse ended−date the abuse began, and age difference was computed as age of accused when the offence began−age of the complainant when the abuse began. Intercoder agreement for each variable is presented in Appendix A.

CODING

Several of the variables listed in Appendix A do not require explanation (e.g., gender of the accused, trier of fact). For other variables (e.g., repression) a more detailed explanation of how the variable was coded is provided below. Notably, we could only code information about a variable if the judge reported the information in his or her decision. There are some details that are legally relevant and that will be included in most or all decisions (e.g., nature of the offence). Other information is less legally relevant (e.g., the qualifications of the person who provided therapy to the complainant) and whether it is reported in a decision will be influenced by a variety of factors, including biases of the judge.

Trial Date

If the case was the trial or the sentencing decision, the trial date was recorded as the judgment date. If the case was from a Court of Appeal (CA) or from the Supreme Court of Canada (SCC) and the trial date was not reported in the judgment, it was recorded as 2 years prior to the CA decision (e.g., a 1997 CA decision was assigned a 1995 trial date) or 3 years prior to the SCC decision: these are rough estimates of the time is could take for a trial decision that is appealed to be released from a CA and from the SCC.

Expert

Only experts who offered social science evidence were recorded (e.g., psychologist, psychiatrist, social worker). If, for instance, a DNA expert testified, he or she was not recorded as an expert in these data.

Description of the Offence

Details of the offence were coded as Levels 1, 2, or 3, from lowest to highest intrusiveness. Appendix B provides a list of the particular acts that constituted each type of abuse. Only the most intrusive form of abuse was recorded.

Frequency

If frequency was specified, the number (e.g., 4 times) or an average of a range of numbers was recorded (e.g., “8 to 10 times” was coded as 9). Sometimes, frequency was described rather than specified. The categories used to code descriptions, as well as the particular descriptors contained in each category, are presented in Appendix B.

Threat

It was reported in some decisions that the child had been threatened, either to refrain from reporting the offence or to submit to the abuse. The nature of the threat was classified as either a threat to the child's (or his/her family, friends, pet) psychological/emotional well-being, to the child's (or his/her family, friends, pet) physical safety, or a threat without further explanation. Appendix B presents a list of the specifics of the threats that were included in each classification.

Dates of the Offence

When possible, the dates that the alleged offence began and ended were recorded from the indictment. If the indictment was not reported, dates were recorded from the facts of the case, where possible. If dates were reported as being one of 2 years (e.g., began in 1976 or 1977 and ended in 1985 or 1986), the most recent dates were used (e.g., 1977 and 1986) to avoid overstating duration and to err in the direction of understating the length of delay.

Repression

Repression was coded as present if it was evident from the text that there was a time when the complainant believed that he/she would not have been able to recall the alleged abuse or that the complainant had “blocked out” critical details that he or she should have known (e.g., the identity of a known perpetrator, for instance a close uncle). All other cases were coded as repression absent. Given the public, legal, and scientific debate that surrounds true and false memories of abuse, and often in the context of “repressed memory,” we chose to categorize cases in terms of the presence or absence of “repression.” Our use of the term is theoretically neutral and simply refers to an understanding that memory for the abuse was not continuously available over time. To illustrate, in the following two cases repression was coded as present: (1) “At age 23, M.L. stated to her therapist that she had always had memories of abuse but she could not ‘put a face to her abuser.’ Then in 1993, on an occasion while hearing from Y.L. that their father had abused her and their half-sister, M.L. stated that the pieces of the puzzle began to come together for her. Memories started to become clearer” [R. v. L.(J.), 1997, at para. 28]; (2) “The crimes were not immediately disclosed by the complainant. She apparently consciously blocked out all memory of the event but suddenly recalled it while under the influence of drugs” [R. v. H.(R.M.), 1990, at para. 3]. Conversely, notwithstanding some forgetting of details, continuous memory was coded in the following two cases: (1) “Although no specific temporal references were provided, it was the evidence of L. B. (1) that there were other instances of the same or similar conduct on the part of the accused involving her which occurred on almost every occasion of her attendance at the home of the accused” [R. v. G. (B.L.), 1996, at para. 17); (2) “There was nothing particularly unique or distinctive about the incidents of sexual abuse that have been described by the complainants. The sexual abuse consisted of numerous acts of sexual intercourse. The Complainants were unable to recall the incidents in any detail.” [R. v. L.(M.), 1998, at para.10].

Therapy

This variable contains information about whether the complainant was reported to have been in therapy. As discussed above, information about therapy was obtained from the text of the decision only: all complainants coded as being in therapy, did seek therapy; however, complainants who were not reported to have been in therapy may have been, in fact, receiving such assistance, despite the judge not recording this fact in his or her decision.

Relationship Between the Accused and the Complainant

The association was coded as parent, other relative, family connection, or community connection. The particular kinds of relationships that comprise each of these classifications are described in Appendix B.

Alcohol

Alcohol was coded as present if the accused was described as having had a drinking problem at the time of the offence(s) or if alcohol was reported to have been a factor in any of the incidents.

Verdict

This variable was coded as a conviction, acquittal, or guilty plea. A conviction included convictions: as charged, of a subset of charges, and of a lesser charge. The case was coded as an acquittal if the accused was acquitted of all sexual charges related to a particular complainant. Guilty pleas included all cases where a trial was not necessary (such cases were sentencing hearings).

Sentence

The sentence could be conditional (i.e., served in the community), probation, incarceration, or a combination of incarceration and probation. In cases involving multiple complainants, the sentences respecting each were recorded in addition to recording whether the multiple sentences were ordered to be served concurrently or consecutively.

Appeal

This includes whether the verdict, sentence, or both were appealed as well as the outcome of the appeal. Depending on the nature of the appeal, an appeal court may uphold the conviction and/or sentence, order a new trial, enter an acquittal, enter a conviction (in Canada, in certain circumstances, the prosecution can appeal an acquittal and/or sentence), or increase/decrease the length of sentence. These categories were classified as: upheld the trial decision, outcome benefited the accused, or outcome benefited the Crown.

RESULTS

The descriptive statistics reported below will not always sum in a predictable way: for instance, the number of sentences reported will not always equal the number of “guilty” pleas plus convictions. This is because there are missing data and the number of empty cells varies across variables. For instance, in 1.3% of the complaints we were unable to identify the accused's plea and in 11% of the complaints we were unable to identify the relationship between the accused and the complainant. When the statistic reported appears too high (e.g., number of outcomes exceeds the number of trials) we explain the discrepancy. In all statistics reported below we present the number of complainants rather than the number of cases.

DESCRIPTIVE STATISTICS

The Legal Context

There were 272 (13%) pre-trial hearings (e.g., application for disclosure/production of counseling records, admissibility of similar fact evidence), 386 (19%) trial decisions, 707 (34%) sentencing hearings, 684 (33%) provincial Court of Appeal decisions, and 15 cases (.7%) from the Supreme Court of Canada. Of the 1800Footnote 2 pleas that were recorded, 571 (32%) were “guilty.” Of the 1145 complaints where the trier of fact was recorded, it was a jury in 495 of the complaints (43%) and a judge alone in the remaining 650 complaints (57%). Differences between jury and judge trials is discussed in detail in Read, Connolly, & Welsh (in press).

There were 748 complaints that involved expert evidence. When an expert was called, it was most commonly to assess the accused (497 complaints, 66% of the experts) and less commonly to provide evidence at trial (251 complaints, 34% of experts). Of the 251 complaints that included a trial expert, 140 (56%), were called by the Crown, 53 (21%) were called by the Defense, and in 58 complaints (23%) an expert was called by both the Crown and the Defense.

Of the 1235 trial verdicts, the majority were convictions: specifically there were 992 convictions (80% of those that went to trial) and only 243 acquittals (20% of trial outcomes). As discussed in Read et al. (in press), this varied considerably as a function of whether the case was heard by a judge (conviction rate of 68.6%) or by a jury (conviction rate of 93.1%). This high conviction rate is consistent with some archival data from the United States (Meyers, Redlich, Goodman, Prizmich & Imwinkelried, 1999). For 1432 complaints (including guilty pleas), we were able to determine the sentence that was passed. In all analyses involving sentence, indeterminate sentences were omitted as outliers. The mean (SD and range in parentheses) number of months on probation, conditional sentence, and incarceration was, 28.11 (11.29, range=3 to 102), 16.25 (7.67, range=1 to 36), 29.34 (31.06, range=1 day to 300 months), respectively. Deciding sentence length involves considering a complex set of factors, one of the most important being severity of the offence (Roberts, 1995; Simon, 1996). The mean (SD and range in parentheses) number of months of probation, conditional sentence, and incarceration for offences defined as Levels 1, 2, and 3 can be seen in Table 1. Although the average length of incarceration increased, there is a great deal of overlap in sentence length between the three levels of intrusiveness suggesting that other factors were important in determining length of incarceration.

Table 1. Length of Sentence as a Function of Level of Intrusiveness of the Offence
Full size table

There were 732 appeals recorded: 322 (44%) from the verdict, 311 (42%) from the sentence, and 99 (14%) from both the conviction and the sentence. Appeal outcome was available in 728Footnote 3 complaints: the trial decision was upheld in 391 complaints (54%); it favored the accused (e.g., order a new trial following a conviction, lower the sentence, enter an acquittal) in 241 of the complaints (33%); and in 96 complaints (13%) the appeal outcome favored the Crown (e.g., increase the length of sentence, order a new trial following an acquittal).

The Complainant

The complainant was female in 76% (n=1537) of the complaints where gender of the complainant was reported (n=2031). She was, on average, 9.37 years (range=1 to 19, SD=3.69) when the abuse began and 12.39 years (range=1 to 28, SD=3.93) when it ended. Put another way, most of the complainants were between about 5- and 13-years old when the alleged abuse began and between about 9- and 16-years old when it ended. The average age of the complainant at trial was 26.09 (range=5 to 64, SD=9.99): The majority of the complainants were early to middle adulthood when they went to trial. The average age difference between the accused and the complainant was 22.98 years (range=1 to 70 years, SD=12.11). In 336 cases (16%), the judge reported that the complainant was receiving or had received some form of therapy. In almost all cases, (n=1941, 94%) the complainant was not reported to have repressed memory for the offence. This is striking in light of the fact that most research on delayed prosecutions of CSA involves claims of repression.

The Accused

There were 1106 accused persons and 2064 complainants: consistent with the possibility that a failure to make a timely official complaint may place other children at risk, there was an average of 1.87 complainants per accused (range=1 to 22). The gender of the accused was identified in 2044 complaints, and the overwhelming majority of accused persons were male (n=2017, 99%). The average age of the accused when the abuse began was 32.82 years (range=11 to 75, SD=12.04), when the abuse ended he was 35.87 years old (range=13 to 76, SD=12.31) and at trial he was, on average, 50.75 years (range=18 to 83, SD=13.87). We identified the connection between the accused and the complainant in 1878 complaints. The most common relationship was one of parent/child (32%, n=602). Other relatives accounted for 25% (n=473) of the relationships; community affiliation represented 21% (n=399) of the complaints; and 19% (n=363) of the relationships were described as family connections. Only 2% (n=41) of the complainants described the accused as a stranger at the time of the offence. If the perpetrator is a stranger to the child at the time of the offence it is more likely that a timely official complaint will be made (Connolly & Read, in press). Moreover, if a timely official complaint is not made, it is unlikely that the perpetrator-stranger would be located many years later.

The Offence

We identified the nature of the alleged abuse in 1814 complaints. Surprisingly, the percentage of cases at levels 1, 2, and 3 were roughly even, 30% (n=549), 35% (n=627), and 35% (n=638), respectively. We expected that more intrusive offences would be more likely to be the subject of delayed prosecution. Frequency of abuse was reported by 1667 complainants: about half (52%, n=866) of the complainants were able to estimate a specific frequency (between 1 and 20) while the others (48%, n=801) described frequency. Most of the complainants (62%, n=535) who were able to specify frequency reported that it had happened once. A further 15% (n=129) of the complainants who specified frequency reported it happened two times. The remaining complainants reported that the abuse occurred between three and 20 times. Among complainants who described frequency, 26% (n=209) described some kind of pattern, 44% (n=356) described the abuse as occurring a lot, while 30% (n=236) reported that the abuse had occurred a few times. Notably, when frequency of abuse was reported it was likely to be repeated abuse (almost 68% of complaints).

The judge reported that some kind of threat accompanied the abuse in 22% (n=454) of the complaints. In 252 (55%) of these cases, the nature of the threat was reported. When the nature of the threat was reported it was most commonly, 62% (n=157), against the physical safety of the complainant or her family. Less frequently (38%, n=95), the complainant's psychological well-being was threatened. The average duration of the offence was 39.77 months (n=1906, range=.03 to 216 months, SD=38.98, median=24 months). When complainants who reported a single episode were excluded, the average duration of offence was a striking 47.80 months (n=1379, range=.25 to 216 months, SD=40.58, median=36 months). On average, complainants waited a very long time to go to trial, 17.13 years (n=1910, range=2 to 49 years, SD=9.51) after the abuse began and 14.03 years (n=1933, range=2 to 48 years, SD=8.67) after it ended.

Multiway Frequency Analyses

A broad description must include a discussion of whether/how characteristics of the offence co-occur. To do this, we completed multiway frequency analyses. These analyses require categorical variables, so variables that are inherently continuous (e.g., age) were converted to quartiles. The analyses of variables associated with the trial and with the accused resulted in contingency tables with far too many expected cell frequencies of zero and less than five to make the results interpretable. Thus, we report analyses of variables related to the offence and to the complainant.

Table 2. Description × Frequency Frequencies (Column Percentages in Parentheses)
Full size table
Table 3. Description × Threat Frequencies (Column Percentages in Parentheses)
Full size table

The first analysis involved variables associated with the offence, specifically description of the offence, offence frequency (frequency was categorized as complaints involving fewer than four instances and complaints involving four or more instances), whether the judge reported the presence of a threat, duration of the alleged abuse, and whether the judge reported that alcohol was involved. Due to missing data, the total number of complaints included in this analysis was 1316. Only 5% of the cells had an expected frequency of less than 5 and none had an expected frequency of 0. The three-, four-, and five-way effects were not significant [χ2 (34)=44.65, p=.11, χ2 (23)=19.90. p=.65, and χ2 (6)=7.79, p=.25, respectively]. However, at least one of the two-way effects was reliable [χ2 (24)=512.77 p < .001] as was at least one of the five possible one-way effects [χ2 (8)=1274.91, p < .001]. We do not report the one-way effects because our interest is if/how variables go together rather than whether frequencies are evenly distributed across the levels of individual variables.

The relationship between description of the offence and frequency [χ2 (2)=8.39, p=.02] can be seen in Table 2: At Level 1 less frequent is more common, at Level 2 the two frequencies are roughly equally distributed and at Level 3 more frequent is more common. Table 3 presents the relationship between description of the offence and threat [χ2 (2)=56.06, p < .001]: As the offence became more intrusive, threats were more likely to have been reported by the judge. As can be seen in Table 4, the relationship between description of the offence and duration was significant [χ2 (6)=50.43, p < .001] because among Level 1 offences, the first and second duration quartiles were the most common; Level 2 offences were most frequent in the second quartile of duration; and the most common duration for Level 3 offences was 61 to 216 months. The trend is for intrusiveness of the offence to increase as duration increases. Table 5 illustrates the relationship between description of the offence and whether the judge reported that alcohol was involved [χ2 (2)=7.04, p=.03]. The percentage of complaints where the judge reported the presence of alcohol was larger for Levels 2 and 3 offences compared to Level 1 offences.

Table 4. Description × Duration Frequencies (Column Percentages in Parentheses)
Full size table
Table 5. Description × Alcohol Frequencies (Column Percentages in Parentheses)
Full size table
Table 6. Frequency × Threat Frequencies (Column Percentages in Parentheses)
Full size table
Table 7. Frequency × Duration Frequencies (Column Percentages in Parentheses)
Full size table
Table 8. Frequency × Alcohol Frequencies (Column Percentages in Parentheses)
Full size table
Table 9. Threat × Duration Frequencies (Column Percentages in Parentheses)
Full size table
Table 10. Threat × Alcohol Frequencies (Column Percentages in Parentheses)
Full size table

Frequency was marginally associated with threats [χ2 (1)=3.142, p=.08], and it was associated with duration [χ2 (3)=188.21, p < .001] and alcohol [χ2 (1)=7.11, p=.01]. As can be seen in Table 6, as frequency increased the proportion of cases where the judge reported the presence of threats also increased. Table 7 illustrates that more frequent abuse was associated with longer duration. The association between frequency and alcohol, illustrated in Table 8, was reliable because, among cases not reported to have involved alcohol, the distribution across the two levels of frequency were relatively even. Conversely, if the judge reported that alcohol was involved in some way, there was a higher percentage of complaints in the low frequency cell than in the high frequency cell.

Threats and duration were associated [χ2 (3)=15.81, p=.001] because, as illustrated in Table 9, threats were more likely to have been reported with complaints of longer rather than shorter abuse duration. Finally, there was a relationship between threat and alcohol, shown in Table 10 [χ2 (1)=3.77, p=.05] because among complaints where the judge reported the presence of a threat, 25.3% also included a report of alcohol whereas in only 15% of the cases where the judge did not report the use of threats was alcohol reported to have been involved. In other words, judicial reports of the presence of threats were associated with more frequent judicial reports of the involvement of alcohol.

The second analysis involved variables associated with the complainant, specifically, gender, age when the alleged offence ended, whether there was a claim of repression, and the relationship between the complainant and the accused. The total number of complaints included in this analysis was 1434. None of the cells had an expected frequency of 0 and 25% of the cells had expected frequencies of less than five. This proportion of cells with an expected frequency of less than five is larger than the recommended level of 20%; however, we proceed because the consequence of violating this guideline is that we lose power (Tabachnik & Fidell, 1989). The four-way effect was not significant [χ2 (9)=5.61, p. =.78]. At least one of the three-way effects [χ2 (24)=52.87, p < .001], one of the two-way effects [χ2 (22)=457.85, p < .001], and one of the one-way effects [χ2 (8)=2249.20, p < .001] was significant. We do not report the one-way effects because our interest is if/how variables go together rather than whether frequencies are evenly distributed across the levels of individual variables. Tables 11 and 12 present frequency tables for each of the significant effects.

The relationship between gender and repression [χ2 (1)=9.068, p=.003] can be seen in Table 11. Although claims of repression were rare indeed, they were more common among females than males. Although the associations between gender and relation [χ2 (3)=253.05, p < .001] as well as age and relation [χ2 (9)=108.85, p < .001] were reliable, meaningful interpretation rests on the association between gender, age, and relation [χ2 (9)=34.11, p < .001]. This three-way association can be seen in Table 12. At all ages, boys were more likely than girls to report abuse by a community member and less likely to claim abuse by a parent and this difference increased as the age of the complainant increased. Indeed, as age increased, girls were more likely to report abuse by a parent and boys were less likely to do so. Moreover, for girls, there was a relatively small increase in allegations of abuse by a community member and only between the first and second age quartiles. However, for boys the increase in allegations of abuse by a community member between the first and second quartile was two-fold and by the last quartile it was a three-fold increase.

Several of the variables associated with the offence and with the complainant were associated with one other variable. It is interesting that more complex three and four-way associations were not found.

DISCUSSION

A primary objective of this paper is to describe criminal prosecutions of HCSA to inform scholarship in two broad areas: research and public policy. To that end, we begin with a an overview of the most frequent characteristic of criminal prosecutions of HCSA followed by a discussion of how these data inform the two broad areas of scholarship.

Table 11. Sex × Claim of Repression Frequencies (Column Percentages in Parentheses)
Full size table
Table 12. Age × Gender × Relation Frequencies (Column Percentages in Parentheses)
Full size table

What does a typical HCSA case look like? The complainant is probably female. On average, she was 9-years old when the abuse began, 12-years old when it ended, and 26-years old at trial. She is unlikely to be reported to have been in therapy, and she is very likely to report continuous memory for the offence. Her abuser is more likely than not to be a male relative and he is, on average, 23-years older than her. On average, he was 33-years old when the abuse began, 36-years old when it ended, and 51-years old at trial. The nature of the abuse is roughly equally distributed across the three levels defined in this report. In the majority of cases, the complainant reports repeated abuse that was sustained over an average period of almost 4 years. A threat is probably not reported to have accompanied the abuse, but if one is reported, it is likely to be against the complainant's or her family's physical safety. How can these data inform discussion in the area of research and development of public policy?

Research Implications

First, with respect to research, scholars have recently begun to study questions that are closely related to the psychological/memory issues raised in these cases, including, memory for trauma in adults (e.g., Christianson & Safer, 1995; Engleberg & Christianson, 2002; Read, 2001; Read & Lindsay, 1997) and memory for childhood trauma (e.g., Prescott et al., 2000; Rivers, 2001) including CSA (e.g., Alpert et al., 1998; Fivush & Edwards, 2004; Ornstein, Ceci, & Loftus, 1998; Williams, 1994). Other relevant research concerns understanding factors that predict/explain delayed disclosure of CSA in community and clinical samples of adults (e.g., Arata, 1998; Badgley, 1984; Finkelhor et al., 1990; Fivush & Edwards, 2004; Lamb & Edgar Smith, 1994; Roesler, 1994; Roesler & Wind, 1994; Smith et al., 2000) as well as in child protection or forensic samples of children who delay reporting CSA (Goodman-Brown et al., 2003; Goodman et al., 1992; Sas & Cunningham, 1995). To the best of our knowledge, this is the only data set of a forensic sample of adult complainants of CSA.

These data provide new entry points for future research to address particular questions facing courts. For instance, notwithstanding that 94% of all complainants reported continuous memory for the abuse, to date, related scholarship has focused on the recovered/false memory debate, to the almost complete exclusion of continuous memory cases (e.g., Connolly & Read, 2003). Also, in spite of the fact that 68% of the complainants claimed repeated abuse, most relevant memory research concerns memory for single occurring events (e.g., Connolly & Read, 2003): research on continuous memory for repeated trauma is needed.

With respect to delays between the end of the alleged abuse and trial, these HCSA data provided surprising results. Specifically, whereas on average a delay of 14 years intervened, fully two-thirds of the complainants were not heard at trial until delays between 5 and 22 years had elapsed and, not infrequently, the delay extended to three and four decades. Investigations of autobiographical events have only rarely approached these kinds of retention intervals and have not usually focused on recollections of childhood events. For those emotional events that have been studied, such as flashbulb events, natural disasters and tragedy, the intervals have rarely exceeded a few years and have been restricted either to adults recalling events experienced as adults or children recalling childhood events. Autobiographical research has occasionally examined very long-term memory for relatively commonplace materials such as academic subjects and high school classmates (for a thorough review of the effects of delay on recollection of autobiographical events, see Read & Connolly, in press). Nonetheless, there is a paucity of empirical research on the retention of emotional or traumatic childhood events recalled after many years. Without such research, we are unable to comment on the likely levels of accuracy of HCSA testimony with any confidence and, as a result, courts are faced with particularly difficult assessments of credibility.

Another area of research that would benefit from these data involves understanding factors that predict which adult survivors of CSA will pursue a legal remedy. The profile presented here could be compared with community and clinical samples of adult survivors who do not file formal complaints. As an example, to study factors that predict delayed official complaints Connolly and Read (in press) compared this profile of HCSA cases with the profile of timely official complaints reported by Badgely (1984). Variables such as a young age of the complainant, more intrusive abuse, reported presence of threats and alcohol, and a closer relationship with the accused were associated with delayed prosecutions. Similarly, the study of factors that predict delayed disclosure may be advanced with these data. As discussed in the introduction, existing scholarship is equivocal in terms of both what predicts delay and the direction of the relationships. Across published studies there is variation in terms of length of delay (from 6 months to several years), definition of sexual abuse (generally self-defined), and verification of the abuse. In the research reported here, the cases involve a delay of at least 2 years, the behavior, if proved, constituted a criminal offence, and the outcome of the trial provided some verification, albeit not indisputable, that sexual abuse had occurred.

Results from the multiway frequency analyses provide some guidance with respect to ongoing research. First, it suggests that more intrusive offences contain multiple indicators of severity including high frequency, threats, long duration, and alcohol. Many of these offence characteristics have been examined individually as possible predictors of delay. These data illustrate that, at least in a forensic sample, several of the factors co-occur. Thus, to study them individually may be misleading. Second, consistent with the possibility that males are more likely than females to be abused outside of the home (e.g., London et al., 2005), being male was associated with a more distant relationship with the accused and this difference was larger among older complainants.

Third, the very strong association between gender and claims of repression raise the question of why females were more than four times more likely than males to claim repression. As reflected in the mutiway frequency analyses, the relationship between gender and repression was not moderated by the complainant's relationship to the accused or the age of the complainant when the abuse ended. Further exploration and understanding of claims of repression may be advanced by careful examination of HCSA cases such as was done in this paper. Court judgments have some distinct advantages over more anecdotal methods of examining cases in which there is a claim that memory was not continuously available to recall. Specifically, descriptions of the relevant events and the complainants' memories of them are elicited in a relatively standardized fashion through both direct and cross-examination with a focus on establishing ground truth. We recognize, of course, that ground truth may not necessarily be the result of this inquiry but, compared to the collection of more anecdotal case descriptions from therapeutic records and other informal sources (see, for example, Schooler, Bendiksen & Ambadar, 1997), we may be better able to place confidence in the facts of a specific case. Indeed, when a complaint is corroborated in some way or its truth proven in court, the presence of repression could be further examined at the same or greater level of detail (i.e., through court transcripts) as that seen in case studies (e.g., Schooler et al., 1997).

Areas of inquiry that are not directly informed by these data, but that are raised because HCSA cases are being heard in criminal courts, as illustrated by these data, involve, for instance, the best way to interview complainants and to gather evidence in these cases (raised by the Home Affairs Committee, 2002) and the perceived credibility of allegations of events that occurred in the distant past. As with CSA trials, credibility is often determinative in HCSA cases (Connolly, Price, & Read, 2006). Although we know a good deal about children's credibility (e.g., Cashmore & Bussey, 1996; Goodman, Golding, Helgeson, Haith, & Michelli, 1987; Leippe & Romanczyk, 1989; Nightingale, 1993; Ross, Jurden, Lindsay, & Keeney, 2003) we simply do not know what report characteristics will lead to a more or less credible report when adults report details of repeated childhood trauma.

Public Policy Implications

In the area of public policy, HCSA cases raise many complex issues that have both legal and social ramifications. For instance, discussions concerning statutes of limitations on sexual offences against children (Home Affairs Committee, Fourth Report, Oct., 2002; Shuman & Smith, 2000) as well as the admissibility of uncorroborated evidence in HCSA prosecutions (e.g., Haber & Haber, 1998; Woodall, 1998) are relevant and timely. These data can both inform the debate and profile the persons most likely to be affected by proposed changes. These data tell us that complainants were very young when the abuse began and were likely to have been abused repeatedly by a person in a position of authority. There may be very long-term mental health consequences of CSA (e.g., Arata, 1998; Browne & Finkelhor, 1986) that preclude pursuit of criminal charges and that complainants are unable to overcome within the limitations period. As Arata (1998) stated, “these assaults that cause the most psychological damage may be the ones that are least likely to be reported” (p. 70). These data also show that in Canada, the average age of complainants in criminal trials is 26 years, on average, 7 or 8 years after complainants reach the age of majority (if the age of majority is 18 or 19). It appears that a substantial period of time beyond the age of majority is required for a typical HCSA victim to come to court. Even in jurisdictions that have statutes of limitations that are tolled until complainants reach the age of majority, unless the limitations period is longer than 7 or 8 years, many complainants may still be barred from proceeding with a criminal complaint.

These data can also inform a debate on the admissibility of evidence. What can we expect a 26-year-old to remember about an event that occurred 17 years earlier when she was only 9-years old. And, perhaps more importantly, what information is expected to be absent from her report? Also, what may render a complainant's memory so unreliable that she should be barred from testifying? Careful analyses of the detail of recall in these cases, as reported by the judge and seen in court transcripts, may provide answers to these kinds of questions. With this information, judges and policy makers will be in a better position to decide evidentiary questions of weight and admissibility (for more detailed discussion and recommendations, see Connolly & Read, 2003; Haber & Haber, 1998).

Public policy also involves allocation of public education resources. In the context of HCSA cases, resources may be allocated to encourage early disclosure. Sorenson and Snow (1991) found that educational programs were an effective catalyst for disclosure among young children. Education may also encourage adults to disclose earlier. Notwithstanding the equivocal findings on clinical repercussions of early disclosure (Arata, 1998; Gomes-Schwartz, Horowitz, & Cardelli, 1990; Goodman-Brown et al., 2003; Lamb & Edgar-Smith, 1994; Paine & Hansen, 2002; Sas et al., 1995; Sauzier, 1989) the forensic implications of a very long delay are less equivocal. In jurisdictions that have an applicable statute of limitations, the effect of waiting could result in a complaint being too old to prosecute. Even without a limitations period, a longer delay increases the chances that any corroborating evidence that might have been available will be lost. It is also possible that the task of judging the credibility of a memory report becomes more complicated as the length of the delay increases (although, as discussed earlier, this is an empirical question worthy of further research). Finally, as illustrated by the very large number of complainants in some trials, the range of complainants per case was one to 20, a longer delay may place additional children at risk of abuse. In short, there are pressing forensic reasons to encourage early disclosure. If educational resources were allocated to encouraging earlier prosecution, where should the resources be targeted? The profile of the “average” complainant may help to make these important public policy decisions.

It is beyond the scope of this paper to present a comprehensive review of all or even most of the questions that can be addressed with these data. Interested readers are referred to Connolly and Read (in press) for an analysis of predictors of delayed official complaints, Connolly, Price, and Read (2006) for an analysis of expert testimony, and Read, Connolly, and Welsh (in press) for analyses of predictors of trial outcome analyzed separately for judge and jury trials.

LIMITATIONS

The present data are subject to all of the limitations associated with archival data. The issues of primary concern are representativeness of HCSA victims and completeness of the archival records. First, the data represent a particular subset of adult survivors of CSA: those who go to court to pursue a criminal charge. The court cases will not include adults who were abused as children by strangers; however, the limitation is relatively unimportant because only a small percentage of CSA victims are abused by strangers (e.g., Badgley, 1984) and earlier disclosure appears to be associated with a more distant relationship between the child and the perpetrator (Arata, 1998; Berliner & Conte, 1990; Di Pietro, Runyan, & Fredrickson, 1997; Sas & Cunningham, 1995; Sjoberg & Lindblad, 2002; Smith et al., 2000).

There are, however, at least two potentially important groups of CSA victims who are not represented in these data; complainants who would have pursued a criminal charge but could not because the Crown chose not to proceed to trial (i.e., exercised prosecutorial discretion) and complainants whose cases were heard in Provincial Courts where the decision was not forwarded to QuicklawTM. This latter loss is particularly important because the majority of criminal cases are heard in provincial courts. Unfortunately, we are simply unable to estimate the size of these groups or to determine if they are different, in predictable ways, from complainants described in these data. Those are empirical questions that we believe are worth pursuing. Nonetheless, the cases included herein are a reasonably comprehensive set of cases that are available to legal professionals and policy makers who will participate in decisions related to research and public policy development.

Second, a judge must decide, based largely on legal relevance, which facts will be recorded in the judgment and which facts will be omitted. We were only able to record information about a variable if relevant information was included in the decision. With respect to variables whose values should have been available because “not present” was not a valid option (e.g., sex, relationship between the complainant and the accused, frequency), this shortfall had its impact on statistical power: if the judge did not provide relevant information, the cell was assigned a missing-data code and excluded from the analyses. However, there are several variables for which we argue that their absence can be assumed even when they were not mentioned in the judgments. These are, specifically, therapy, repression, threats, alcohol, and experts. It is unlikely that a judge would explicitly report the absence of these factors (e.g., “this complainant did not repress memory for the abuse” or “this complainant was not in therapy”). Thus, if the factor was not mentioned, it was coded as not present (e.g., no repression, no therapy) and included in the analyses. Of course, it is possible that the judge failed to mention the factor because it was deemed irrelevant. We submit that this is unlikely for the repress, threat, and expert variables because they would be deemed legally relevant. However, it may be that some complainants we have recorded as not being in therapy were receiving or had received some form of counseling and that alcohol was a factor in some cases but not mentioned by the judge.

It may be considered a shortcoming of these data that they include only Canadian criminal cases. In the introduction we argued that HCSA prosecutions are generating debate and discussion internationally: they are not a uniquely Canadian phenomena. We have no reason to believe that case profiles are different outside of Canada, although that is an empirical question. Even if there are unique national characteristics in terms of the case profiles, the issues faced by the courts will be similar across countries: how to evaluate the credibility of a memory report of something that happened, probably repeatedly, a long time ago, when the complainant was a child.

In conclusion, most children who experience CSA do not report the abuse immediately, and many do not disclose the abuse until adulthood. Some will seek a legal remedy during adulthood. Although the phenomenon is international, the range of available remedies is shaped by regional/national and legal policy. In several jurisdictions that do not have criminal statutes of limitations, criminal courts are often charged with the very difficult task of adjudicating allegations of events that happened, often repeatedly, many years or decades earlier. These cases raise numerous questions that psychologists are uniquely qualified to address. Notwithstanding this expertise, very little research is available on the issues. It is our hope that a better understanding of the phenomenon will inform advancement in the areas of research and the development of public policy.