Interpersonal and systemic aspects of emotional abuse at work

The research of Loraleigh Keashley, Ph.D.
Academic Director, MA in Dispute Resolution
College of Urban, Labor, and Metropolitan Affairs
Wayne State University, Detroit, USA

Interpersonal and systemic aspects of emotional abuse at work:

The target's perspective

"The new profile of violence at work which emerges is one which gives equal emphasis to physical and psychological behaviour, and one which gives full recognition to the significance of minor acts of violence." (Chappell & Di Martino, 1998; pg. 12)

This quote from the International Labor Organization's report on violence at work succinctly captures what we as researchers have come to realize and employees and employers have known for a long time; that workplace violence is not just physical. Indeed, evidence is accumulating that nonphysical or more psychological forms of aggressive behaviors are more frequent than physical assaults and equally if not more devastating for employees (both as targets and witnesses), their supporters (family, friends) and organizations than physical assaults (NNLI, 1993; Baron & Neuman, 1996; Keashly, 1998; Richman et al, 1999). In terms of this "other face" of violence, sexual harassment has been the focus of much research and policy attention and we know the range and impact of its effects (e.g., Barling et al, 1993; Fitzgerald et al, 1997). More generalized workplace harassment which is not tied to group membership and is known by many names including bullying, mobbing, and emotional abuse, is now the focus of more research (Einarssen, 1999; Hoel, Rayner, & Cooper, 1999; Keashly, 1998; Richman et al, 1999) and policy (Yamada, 2000) attention. Examples of these behaviors include exclusion from key work activities, silent treatment, putdown in front of others, yelling and screaming, flaunting status, and taking credit for work

Interestingly, with regards to the impact of these different forms of violence, the range and nature of the effects are remarkably similar. In a comparative review of the physical violence, sexual harassment, and emotional abuse literatures, Rogers (1998) noted that while there are outcomes unique to each form (e.g., physical violence often involves bodily injuries such as broken bones and bruises; sexual harassment can affect the victim's view of and behavior toward members of the opposite sex), there appears to exist a set of core outcomes associated with all of these forms of violence. Specifically, whether physically assaulted, sexually harassed, or emotionally abused, victims seem to experience immediate reactions such as anger, humiliation, fear, and/or tension; physical symptoms such as nausea, headaches, sleep difficulties, and/or weight loss/gain; emotional ailments such as decreased levels of self-esteem and/or increased levels of anxiety or depression; and organizational outcomes such as increased intentions to leave the job and/or decreased levels of job satisfaction, organizational commitment, and/or morale. A similar observation has been made in the traumatic stress literature. Newman (1987) notes that despite the differing personalities of the victims and the differing circumstances surrounding the events, the victims share remarkably similar emotional reactions and psychological defenses.

Thus, both in terms of frequency of occurrence in people's daily working lives and the nature and range of effects, we need to define and assess the domain of workplace violence broadly to include not just physical events but also verbal and the more psychological events. Practically, this means we need to examine these different behaviors concurrently as opposed to separately as has occurred in much of the organizational literature to date (ex., Neuman & Baron, 1997; O'Leary-Kelly, Griffin, & Glew, 1996; Robinson & Bennett, 1995; Silver & Glicken, 1990). One of the goals of the current study was to assess the relative frequencies of these different types of violent behaviors by having people indicate the extent to which they were on the receiving end of a variety of behaviors within the span of a work week.

The second goal of this study concerns the similar effects of these very different forms of violence. This is of particular interest as sexual harassment and emotional abuse have been characterized as "low-level" violence in comparison to physical assaults (e.g., Arnetz, 1998). If this is an accurate description then it would be expected that low level (i.e., less severe) violence should have less severe effects. An occupational stress framework would suggest that the explanation may lie in the subjective evaluation of the experience by the recipient. Lazarus and Folkman (1984) argue that it is not the occurrence of a "stressful" incident which best predicts short and long term effects, but how the incident is appraised by the victim.. The individual must determine whether the incident is meaningful or salient (i.e., positive, negative, or irrelevant in its implications for well-being; primary appraisal) and if there is anything that can be done to deal with the incident(i.e., perceived control or secondary appraisal).An event is more likely to be seen as a threat when the event is perceived as negative and difficult to control. In addition, Lazarus and Folkman (1984) suggest that certain contextual variables (e.g., novelty, predictability, and duration) influence whether an events is appraised as threatening, or stressful. Thus, in this study, we examined how incidents of physical, sexual, and emotional violence at work are appraised by recipients. This included identifying certain contextual variables which would predict incident appraisal and examining the relationship between incident appraisal and incident stressfulness.

Forms of violent behavior

The first area of inquiry in this study concerned whether people differ in their appraisals of physical, sexual, and emotional incidents of workplace violence. Since few studies have empirically examined these three forms of aggression concurrently, there was little guidance in predicting how these three forms would be appraised. Roger's (1998) earlier work found that the three types of behavior differed in terms of perceived severity, with physical behaviors seen as most severe followed by sexual and emotional behaviors respectively. Thus, it was hypothesized in the current study that the physical incidents would be perceived as the most meaningful incidents (primary appraisal) followed by the sexual and emotional incidents, respectively. Regarding secondary appraisal or perceptions of control, the existence of legislation and organizational policies addressing physical violence and sexual harassment in the workplace and the lack of policies regarding emotional abuse (Keashly, 1998; Yamada, 2000) have implications here. It is likely that employees feel that they have support and avenues for action, and hence more control over incidents of physical and sexual violence at work. Thus, it was hypothesized that emotional incidents would be perceived as more difficult to deal with than incidents of physical violence or sexual harassment.

Contextual variables

Understanding why physical, sexual, and emotional behaviors may be appraised differently requires consideration of the context within which the behavior occurs. For example, the sexual harassment literature recognizes that potentially harassing behaviors cannot be identified or examined in isolation of their context (York, 1989).Extending this reasoning to the other forms of violence, we hypothesized that these aggressive behaviors would be appraised differently due to contextual factors which were independent of the form of behavior. That is, once differences in these factors were controlled for, any differences in cognitive appraisal of these different violent incidents would disappear.

Following the sexual harassment research on appraisal, the factors examined included characteristics of the actor (e.g., position, sex), prior history with the actor, and attributional processes (e.g., intent).

Characteristics of the actor

Two actor characteristics were examined in this study: organizational position and sex. Regarding organizational position, it was hypothesized that behavior of a higher status actor (supervisor or coworker) would be appraised as more meaningful than the behaviors of lower-status actors (subordinates or clients). Sheffey and Tindale (1992) found that incidents were more likely to be perceived as sexual harassment when the actor was of a higher status than the target. With respect to verbal abuse of nurses and medical students, abusive actors are often distinguished in terms of their position (Baldwin et al, 1991).

In addition, we hypothesized that behaviors of higher status actors would be perceived as more difficult to control than those of lower status actors. Higher status actors have more organizational power or resources and hence are more likely to be able to follow through on threats and to be able to influence perceptions of the target inside the organization. As a result, targets may perceive that they have fewer resources to draw on in order to control incidents from higher status actors and that actions to control these incidents may be associated with greater personal consequences.

In terms of actor sex, research in sexual harassment suggests that behaviors performed by male actors toward female targets are perceived as being more frequent and serious (Fitzgeral & Shullman, 1993; Gutek, 1985). Results on the effect of actor sex on perceptions of emotional and physical forms of aggression are inconsistent (Keashly et al, 1994; Price Spratlan, 1995; Whittington et al, 1996). To the extent that sex is associated with power and that incidents perpetrated by powerful actors are perceived as more abusive, it could be argued that incidents perpetrated by men would be perceived as more abusive than the same incidents perpetrated by women. Thus, we hypothesized that incidents perpetrated by men would be perceived as more meaningful (primary appraisal) than those perpetrated by women. Further, based on the same power argument made with actor position, it was hypothesized that incidents perpetrated by men would be perceived as more difficult to control (secondary appraisal) than incidents perpetrated by women.

History with the actor

 The importance of previous history in perceptions of abuse has been suggested by several researchers. First, many researchers in the emotional abuse and sexual harassment areas incorporate this construct in their definitions of abuse/harassment (see Fitzgerald & Shullman, 1993; Keashly, 1998). That is, sexual harassment and emotional abuse are often defined as patterns of behaviour which involve a series of repeated events. Second, it has been suggested that if an actor makes sexual overtures toward a target consistently over time, the behaviour is more likely viewed as sexually harassing (Pryor, 1985). It is generally believed that this is because behaviour exhibited by the actor over time is causally attributed to the actor's character. Thus, it was concluded that history with the actor is an important variable in predicting how an incident is appraised by victims regardless of the form of the incident. Sexual harassment research suggests that previous history with the actor (i.e., increased actor consistency) is positively related to the perceived stressfulness of an incident. According to Lazarus and Folkman (1984), the more meaningful an incident, the more stressful it will be appraised. Thus, it was hypothesized that prior history with the actor would be associated with higher appraisals of meaningfulness. There is little research to suggest how history with the actor may influence perceptions of control. Hence it was hypothesized that history with the actor would have its influence on cognitive appraisal through the primary appraisal variables.

Incident Severity

Paterson and Neufeld's (1989) work on the parameters of stressful situations suggests that greater stimulus severity is associated with appraisals of greater threat. The sexual harassment literature also suggests that incidents that are perceived as more extreme or explicit will be more likely to be assessed as harassment (Fitzgerald & Shullman, 1993). Incident severity was also identified as an important variable in the perception of incidents of violence in an earlier study (Rogers, 1998). Thus, it was hypothesized that the greater the perceived severity of an incident the more meaningful the incident would be perceived. Furthermore, it was felt that perceptions of control would be influenced by stressor severity. To the degree that incidents which are perceived as more severe are experienced as more overpowering, the targets of these incidents may feel less able to control these incidents. These more severe incidents may in some way immobilize targets and in turn influence their ability to control the incident (See Keashly, in press). Hence, it was hypothesized that the higher the perceived severity of an incident the more difficult the incident would be to control.

Attributional Variable: Intent

 Negative intent has been demonstrated as a key predictor for satisfaction in interpersonal interactions. For example in their review of the marital satisfaction literature, Bradbury and Fincham (1990) noted that events are seen and experienced as more negative when the partner is considered to have behaved with negative intent (i.e., to deliberately hurt their partner). Regarding workplace interactions, Pryor and Day (1988) found that people are more likely to judge an actor's sexual overtures toward a target as sexually harassing if these behaviours are attributed to the actor's negative intentions. It seems probable that perceptions of intent to harm would be important in the appraisal of all types of incidents, regardless of their form. For this reason it was hypothesized that if the target perceives the actor to have intended to harm him/her with the behavior, the target is more likely to perceive the behavior as meaningful. With little research linking intention to harm with perceptions of control, it is not clear how such an assessment may influence the victim's ability to control the event, or their resources to do so. As a result, it was hypothesized that intent to harm would have its influence through the primary appraisal variables

Cognitive appraisal and stressfulness

Lazarus and Folkman's (1984) model suggests that primary and secondary appraisal mediate the relationship between the antecedent variables and the measures of distress. There are numerous studies which substantiate the relationship between cognitive appraisal and measures of distress (Elliott et al., 1994; Lazarus et al, 1986; Solomon et al., 1989; Vinokur et al., 1990).With regard to violent incidents, it was hypothesized that the primary appraisal variables would be positively associated with the stressfulness of an incident, and the secondary appraisal variables would be negatively associated with the stressfulness of an incident.

Control Variable: Trait Negative Affectivity

The relationship between self-report measures of stress and strain can be inflated by negative affectivity since people who are high in negative affectivity are prone to report more job related distress and dissatisfaction compared to those low in negative affectivity (Brief, Burke, George, Robinson, & Webster, 1988). The current study relied on self-report survey data, and hence was threatened by the mono-source bias described by Brief et al. (1988). Further, since the participants in the present study were chosen because of their likely exposure to workplace violence, negative affectivity presented itself as a potential confound(Barling et al., 1993; Schneider et al., 1997;Rogers & Kelloway, 1997). Thus, negative affectivity was explicitly measured and controlled for in this study. It was expected that the hypothesized relationships would exist after controlling for negative affectivity.

The current study

In order to address the research questions specified above it was necessary to sample from an occupational setting in which employees are potentially exposed to all three forms of violence, perpetrated by a variety of sources. After an examination of the literature, it was concluded that health care settings fulfil both of these requirements. First, there is substantial documentation that health care workers (e.g., nurses, therapists) experience physical, sexual, and emotional forms of violence during their daily work routine (e.g., Cox, 1990; Diaz & McMillin, 1991; Graydon et al., 1994; Lanza, 1983). Second, the research suggests that this group of workers is mistreated by individuals in a variety of occupational positions, including supervisors, coworkers, physicians, and patients/clients. For these reasons, employees in health care settings served as the participants in this study.

Method

Participants

Participants were 236 employees from four health care settings in Ontario: a small rural general hospital (N=33), a medium (N=31) and a large (N=168) mental health hospital, and a small (N=4) mental health clinic in a large urban center.

Of the 236 participants, 9 participants were excluded because they did not experience any incidents of violence during the study, nor could they recall any past incidents. The final sample was 227 with 171 females and 56 males. The mean age of participants was 43 years (range = 22 - 64 years), and their mean tenure at their institution was 12 years (range = 1 month - 37 years). Eighty-six participants (38%) were nurses, 58 (25%) were other clinical staff (e.g., psychologists, psychiatrists, social workers, physiotherapists, recreational therapists, etc.), 27 (12%) were managers, 25 (11%) were professional/technical staff (e.g., pharmacist, vocational workers, etc.), and 31 (14%) were support staff (e.g., maintenance, clerical, cleaning, etc.). Further, 195 (86%) of participants worked full-time for their institution, with the remainder working either part-time or on a contract basis. Since 86% of participants worked in mental health settings, the majority of patients that these participants came into contact with were patients with mental health challenges.

Materials

Workplace Violence

The fifty-two item Workplace Violence Checklist (WVC) assessed the occurrence/frequency of numerous incidents of workplace violence (emotional, physical, and sexual incidents) over the span of a working day. The development of this checklist occurred in two phases. In the first phase, a list of 238 incidents was compiled based on a review of relevant literatures (e.g., sexual harassment, emotional abuse, physical violence, workplace deviance, family violence and occupational stress). These incidents were then categorized into the three forms and redundant and occupation-specific items were removed. The resultant WVC includes 15 physical items (e.g., "been hit with an object), 24 emotional items (e.g, "yelled or shouted at"), and 13 sexual items (e.g., "told suggestive/offensive stories or jokes"). For each item the subject was asked how often the item occurred at work that day using a four-point scale (0=never, 3=4 or more times). Researchers have suggested that measures of internal consistency are not meaningful nor appropriate for causal indexes (versus effects indicators; Bollen & Lennox, 1991). For this reason, the reliability of this index is not reported here.

Violent Incident Description

Of the experiences documented on the WVC, each participant described in detail one or two incidents. Some of these experiences of violence involved only one item from the checklist while others involved multiple items. Each incident was coded for the form of violence (i.e., emotional, physical, or sexual) involved in the incident.

Contextual Variables

Intent to Harm

 Intent to harm was assessed through one item developed for the purposes of this research. This item, "this individual intended to harm me with this incident," was rated on a five-point scale (1= strongly disagree, 5= strongly agree). Using a single item to measure a psychological construct is often considered problematic since measures of internal consistency cannot be generated for these measures (Wanous & Reichers, 1996). However, if the construct being measured is narrow and unambiguous to the participant, many researchers believe that a single item is acceptable (Sackett & Larson, 1990; Wanous & Reichers, 1996). The construct of intent to harm was judged to be sufficiently narrow to warrant a single item in this study.

Incident Severity

 The severity of the incident was based on the severity ratings derived from an earlier study (Rogers, 1998). Based on the qualitative description of each incident provided by the participants in the present study, the researcher listed the items from the checklist which were involved in the incident. The researcher then chose the item with the highest mean severity rating based on the ratings from Rogers (1998) study. This rating represented the perceived severity for that incident. The mean severity ratings for the 52 items in the checklist ranged from 3.59 to 6.86 (on a 7 point scale) with higher ratings representing greater severity.

Actor Status and Sex

Both actor sex and actor status (supervisor, coworker, subordinate, patient/client) were gathered as part of the participants' qualitative description of each incident. Actor sex was coded 1 for females and -1 for males. Actor status involved two contrasts: 1) patient group vs. others and 2) supervisor vs. coworker.

History with the Actor

Past history with the actor was assessed through three items developed for this study. Two of these items were positively keyed (i.e., "this individual has treated me in this exact way before"; "this individual has treated me in related ways before"), and one item was reverse keyed (i.e., "this individual has not treated me in ways similar to this in the past"). Items were rated on a five-point scale (1=strongly disagree, 5=strongly agree). The internal consistency of this scale was acceptable (alpha=.88).

Primary Appraisal

Centrality

 The centrality sub-scale from Peacock and Wong's (1990) Stress Appraisal Measure (SAM) was chosen. This 4 item sub-scale assessed the implications of an incident for one's well being. Items were rated on a five-point scale (1=not at all, 5=extremely). This scale demonstrated strong internal consistency with an alpha of.89.

Importance

The salience sub-scale from Vitaliano, Russo, Weber, and Celum's (1993) Dimensions of Stress scale was selected and slightly modified (e.g., items were put into the past tense). This 4 item scale assessed the personal relevance or importance of an incident. Each item was rated on a five-point scale (1=strongly disagree, 5=strongly agree). Internal consistency was excellent (alpha=.90).

Secondary Appraisal

Control-Self

 The control-by-self scale from Peacock and Wong's (1990) SAM was used to measure the extent to which the situation was controllable by the participant him or herself. This scale consisted of four items rated on a five-point scale (1=not at all, 5 extremely). Items were revised to be in the past tense. This scale demonstrated high internal consistency with an alpha of .85.

Control-Others.

The control-by-others scale from Peacock and Wong's (1990) SAM was used to assess the extent to which a situation was controllable by other people, that is, whether the individual had access to people who could help him/her deal with the situation. This scale consisted of four items rated on a five-point scale (1=not at all, 5 extremely). Items were revised to be in the past tense. This scale demonstrated strong internal consistency with an alpha of .88.

Barriers

 This scale was created for this study and assessed perceived barriers internal to the organization which may have prevented the individual from responding to the incident in the way he/she would have preferred. Even if the individual did have access to the resources detailed in the Control-Self and Control-Others scales, they may have chosen not to act on them due to other factors. The barriers scale consisted of four items rated on a five point scale (1=strongly disagree, 5=strongly agree; e.g., "were the costs associated with dealing with this incident greater than the benefits?"; "did the hospital make it difficult for me to deal with this incident?".The items in this scale were adapted from the work of Folkman and Lazarus (Folkman et al., 1986a; Folkman et al., 1986b) and Dewe (1991; 1992).

It was predicted that the four items which comprise the barriers scale would load on a factor separate from the other two secondary appraisal measures (control-self and control-others). A factor analysis including all of the items in the secondary appraisal measures supported the four-item measure of barriers. Further, this scale demonstrated satisfactory internal consistency (alpha=.77).

Incident Stressfulness

The stressfulness sub-scale of Peacock and Wong's (1990) SAM was used as a global measure of incident stressfulness. This scale consisted of four items revised to place them in the past tense and rated on a five-point scale (1=not at all, 5=extremely). Internal consistency was high (alpha=.88).

Control Variable

Negative Affectivity

 Watson, Clark, and Tellegen's (1988 ) Positive and Negative Affect Scales were used in this study. The scale consists of 10 adjectives which describe mood and are rated on a five-point scale (1=very slightly, 5=extremely). Instructions were altered to emphasize mood in general (trait negative affectivity) rather than current mood (state negative affectivity). Internal reliability was high (alpha=.84).

Procedure

Participants completed three surveys: 1) demographic and affectivity variables; 2) Workplace Violence Checklists; and 3) description and rating of selected incidents. At the first data collection session, participants filled out a ticket for a lottery and completed the consent form. The first survey assessed participants' affectivity and demographics (i.e., tenure, work area, work status, age, sex, and education). Once the participant completed this survey, they were provided with the second survey which contained five copies of the Workplace Violence Checklist. It was explained that each checklist would ask the participant whether certain incidents occurred over the span of that work day and that they were to complete one checklist at the end of each working day for five days. A time was then arranged with the participant to meet five work days later for the second data collection session.

In the second session, the participant submitted the checklists for the week. If two or more items occurred over the last five days, the participant was asked to identify the most "memorable" incident documented on the checklists. He/she was then asked to briefly describe the incident. This description included: (1) what lead up to the incident; (2) what occurred during the incident; (3) the relationship between the participant and the actor(s) (i.e., co-worker, supervisor, subordinate, patient/client); (4) the sex of the actor(s); and (5) where the incident occurred (i.e., in the hospital or in the community). Next the participant was asked to rate this incident on several scales (i..e, appraisal, context-related, and stressfulness scales). Once these ratings were completed, the participant was asked to describe a second incident from the checklist ("researcher selected" incident). Once this incident was documented the participant was asked to rate this "researcher selected" incident on the same scales as the first incident.

If the participant had only checked off one item on the checklist he/she was asked to describe the one incident and rate it on the scales. Then, the participant was asked whether any of the items had occurred over the last six months to a year. Those participants who could recall a "past" incident were asked to describe the incident and rate it on the scales. Those who could not recall any past incidents rated only one incident. Participants were then debriefed.

Each incident was coded for the sex of the actor, the status of the actor (i.e., coworker, supervisor, patient/client, or subordinate), the behavioral items involved in the incident (maximum of seven), the severity of the incident, and the form of violence involved in the incident. Fifteen percent of the incidents recorded were   by a trained second party in order to assess inter-rater reliability. Cohen's kappa was used as the reliability measure (Cohen, 1960). Kappas ranged from.91 to .95, with an average Kappa across coding categories of .93.

Results

Rationale for Analyses

The hypotheses were examined through the use of multivariate analysis of variance (MANOVA), multiple regression (MR), and canonical correlation analyses. The MANOVAs used Pillai's criterion to assess the overall Fs. Further, significant overall Fs resulting from the MANOVA analyses were examined through discriminant analyses and centroids. Since the DVs (i.e, the cognitive appraisal variables) in these analyses were conceptually related to one another, a follow-up procedure which examined differences between the three levels of form of violence on the primary and secondary appraisal variables as a group was seen as most appropriate (Klecka, 1980).

Descriptive characteristics

Examination of the correlations in Table 1 reveals that the five appraisal variables (i.e., importance, centrality, control-others, control-self, and barriers) were all significantly correlated with stressfulness. As well, the two primary appraisal variables (centrality and importance) were significantly correlated as were the three secondary appraisal variables (control-others, control-self, and barriers).

Table 2 presents the percent of participants who reported experiencing each of the incidents of violence over the five day period. These percentages were calculated by collapsing responses into two categories of "never occurring" and "occurring at least once".. Of the ten most frequently reported incidents, 7 were emotional, 2 were sexual and one was physical in form. Participants identified 8 additional incidents which were not on the original WVC. Five were emotional in form: 1) not given access to needed resources/information; 2) had work sabotaged by others; 3) told offensive stories of a nonsexual nature; 4) exposed to verbal mistreatment of others (witnessing); and 5) criticized and belittled in private. Two items were physical in form: 1) followed or chased and 2) been tripped. The final item was sexual in form: target of inappropriate/suggestive gestures.

Of the 227 participants, 187 discussed two incidents and 40 discussed one incident for a total of 412 incidents. Of the total incidents described in detail, 55% were emotional, 26% were physical, and 19% were sexual in form. The emotional incidents involved only emotional items since the definition of emotional violence excluded any physical or sexual items. Of the physical incidents, 23% of these incidents were exclusively physical, whereas 77% of these incidents involved both physically- and emotionally-related items. Of the sexual incidents, 90% were exclusively sexual, while 10% involved physical and/or emotional items. Hence it would appear that both incidents of emotional and sexual violence can, and often do, occur in isolation while incidents of physical violence often involve emotional violence.

Based on these 412 incidents, incidents of emotional violence were most frequently perpetrated by coworkers (41%), followed by clients (29%) and then supervisors (24%). Further, emotional incidents were most often perpetrated by female actors (62%). Incidents of sexual violence were most frequently perpetrated by clients (67%), followed by coworkers (29%), with most actors being male (83%). Last, physical incidents were overwhelmingly perpetrated by clients (94%), with most actors being male (67%).

Tests of Proposed Relationships

Form of violence and cognitive appraisal

 In order to test the hypotheses that incidents of physical, sexual, and emotional violence were appraised differently by participants, a MANOVA was run examining form of violence on appraisal. Appraisal was assessed through centrality and importance (primary appraisal) and control-others, control-self, and barriers (secondary appraisal). Four cases were removed due to being outliers. The ensuing analyses were based on a sample of 223 (emotional group=79, physical group=74, and sexual group=70). The combined DVs were significantly affected by the three levels of form, approximate F (10, 434)=3.52, p <.001. This association between the DVs and form of violence was small, with eta 2 =.14.

The discriminant analysis derived two canonical discriminant functions. Examination of the rotated correlations between the first canonical function and the dependent variables defined a dimension of meaningfulness, or primary appraisal, with importance ( r =.84) and centrality ( r =.93). The three levels of form differed on this dimension of meaningfulness with the emotional incidents scoring highest (centroid=5.09) followed by the physical incidents (centroid=4.86) and the sexual incidents (centroid=4.11). Tukey tests were run to determine whether these centroids were significantly different from one another. While the emotional and the physical incidents were significantly more meaningful than the sexual incidents ( p <.01), the physical and emotional incidents did not differ. Hence, there was partial support for the original hypothesis since the emotional and physical incidents were perceived as more meaningful than the sexual incidents.

Examination of the rotated correlations between the second canonical function and the dependent variables defined a dimension which reflected the ability to manifest control, or secondary appraisal, with control-others ( r =.79), control-self ( r =.41), and barriers ( r =-.77). The three levels of form differed on this dimension of ability to manifest control with the physical incidents scoring highest (centroid=7.85) followed by sexual incidents (centroid=7.28) and the emotional incidents (centroid=6.93). Tukey tests suggest that the physical incidents were significantly easier to control than the emotional incidents ( p <.01) but that the sexual incidents did not differ significantly from either the emotional or the physical incidents. Thus as hypothesized, the levels of secondary appraisal did differ by form of violence such that the emotional incidents were perceived as more difficult to control than the physical incidents.

In summary, the physical and emotional violence incidents were perceived as the most meaningful incidents, followed by the sexual incidents. Further, the emotional incidents were perceived as the most difficult with which to manifest control. Along with being among the most meaningful incidents, the physical incidents were perceived as the easiest with which to manifest control. Last, the sexual incidents were the least meaningful and were between the emotional and physical incidents on ability to manifest control.

Contextual variables and cognitive appraisal

 It was hypothesized that the contextual variables examined in this study would be related to the primary and secondary appraisal variables, and that these variables would explain the relationship which exists between form of violence and cognitive appraisal. This hypothesis was assessed through two separate analyses. The first analysis was a MANCOVA, with form of violence on the appraisal variables. Adjustment was made for seven covariates: negative affectivity, history with actor, intent to harm, severity, sex of actor, and status of actor. Sample size for this analysis was 206 with the deletion of 17 cases for missing data. As hypothesized, the combined DVs were significantly related to the combined covariates, approximate F (35, 980)=3.17, p <.001.

Effects of form of violence on the DVs after adjustment for coveariates were then investigated. Form of violence was not significantly related to the combined DVs. Hence, as hypothesized, with the inclusion of the covariates the previous significant relationship between form and appraisal was no longer significant.

Since the cognitive appraisal variables are significantly correlated these relationships were addressed through canonical correlation, specifically using the OVERALS method. The first set of variables in this analysis were the contextual variables: negative affectivity, intent to harm, history with actor, status of the actor, actor sex, and severity of incident. The second set included importance and centrality (i.e., primary appraisal). The third set included barriers, control-others, and control-self (i.e., secondary appraisal). A model was tested using one- and two-dimensional solutions. The FIT of the two- dimensional solution was .56 while the FIT of the one-dimensional solution was .64. The one-dimensional solution was accepted due to its superior goodness-of-fit and its parsimonious nature (see Table 3). The importance of the variables in each set is assessed through the size of the loading relative to each other, while the magnitude of the loadings suggest which variables should be examined (Rosin & Korabik, 1991). A cutoff value of .40 (or 20% overlapping variance) was established by the researcher for interpreting loadings (Comrey, 1973).

Two of the context variables made important contributions to the canonical correlation: intent to harm (.61) and status of the actor (.40). Both of the primary appraisal variables, importance (.72) and centrality (.73), were found to be important. From the set of secondary appraisal variables barriers (.78) and control-self (-.48) were found to make an important contribution.

Thus, consistent with the hypothesis, those incidents in which the actor was perceived as intending to harm the victim were perceived by the victim as more important and more central, and were associated with lower levels of personal control and with greater barriers. In order to examine the relationship between actor status and primary and secondary appraisal the centroid plots were examined. This examination suggested that those incidents perpetrated by supervisors (centroid=1.06) were associated with the highest level of importance, centrality, and barriers, and with the lowest levels of personal control. The actors associated with the next highest levels of importance, centrality, and barriers, and the next lowest level of control were coworkers (centroid=.44), followed by subordinates (centroid=-.07), and finally patients (centroid=-.29). Hence incidents perpetrated by patients were associated with the lowest levels of importance, centrality, and barriers, and the highest levels of personal control. These results partially support our hypotheses since certain context variables, but not all, were related to cognitive appraisal.

Mediating role of cognitive appraisals

According to Baron and Kenny (1986), a variable functions as a mediator when, assuming the IV is significantly related to the DV: (a) the IV explains a significant amount of variance in the presumed mediator; (b) the presumed mediator explains a significant amount of variance in the DV, and (c) the relationship between the IV and the DV is no longer significant, or drops, when the mediator is controlled for. In each multiple regression, negative affectivity served as an IV and was entered on its own in the first block to control for its effects on these relationships (Brief et al., 1988). Intent to harm and actor status were the other predictors because of their relationships to the primary and secondary appraisal (described above). Stressfulness was the criterion. The mediators were importance, centrality, control-self, and barriers because they were related to the contextual variables.

1. Intent to harm and stressfulness

 First, intent to harm predicted a significant amount of variance ( R 2 =.23) in stressfulness, F (2, 217)=32.56, p <.001. Second, intent to harm explained a significant amount of unique variance in all four appraisal variables, namely importance; F (2, 217)=20.82, p <.001, centrality F (2, 217)=26.36, p <.001, control-self F (2, 217)=4.65, p <.01, and barriers F (2, 217)=9.08, p <.001 (see Table 4).

Third, the four mediators in this study predicted a significant amount of variance in stressfulness, F (5, 216)=89.43, p <.001. However only importance, centrality, and barriers predicted a unique amount of variance in stressfulness (see Table 5). For this reason, control-self was dropped as a potential mediator.

Finally, to meet the last requirement for mediation the relationship between the IV and the DV must decrease significantly when the mediator is controlled for. Intent to harm still explained a significant amount of variance in stressfulness when importance, centrality, and barriers were entered into the equation (see Table 5). These results suggest that the appraisal variables did not fully mediate the relationship between intent to harm and stressfulness. Given this, the appraisal variables still may have partially mediated this relationship. The significance of the reduction in beta between the direct path from intent to stressfulness and the indirect paths through the three mediators was examined using Goodman's method (as cited in Baron & Kenny, 1986). In line with mediation, the decrease in beta was significant for all three mediators, namely importance, z =3.82, p <.01, centrality, z =6.05, p <.01, and barriers, z =2.84, p <.01. These results are consistent with partial mediation. That is, it would appear that the amount of variance explained in stressfulness by intent to harm was partly shared with the appraisal variables and was partly unique.

2. Actor status and stressfulness

 Actor status was assessed through two contrasts: (1) patients versus other actors, and (2) supervisor versus coworkers. Following the same procedure as above, the first MR examined the relationship between actor status and the appraisal variables (i.e., importance, centrality, control-self, and barriers). Actor status explained a significant amount of variance in barriers, R 2 = .17, F (3, 212)=16.50, p <.001, but did not explain a significant amount of variance in the other three appraisal variables. The one contrast which explained the relationship was the patients versus other actor contrast ( Beta = -.46, t = -6.50, p <.01). Since this contrast was not significantly related to stressfulness (see Table 1), mediation was not a possibility. Instead, the patient versus other actors contrast appeared to explain variance in the barriers variable which was not shared by the stressfulness variable. This finding goes counter to the hypothesis that cognitive appraisal mediates the relationship between actor status and stressfulness.

In sum, the relationship between intent to harm and stressfulness was consistent with partial mediation such that part of this relationship was direct in nature and part of it was mediated through the three appraisal variables (i.e., importance, centrality, and barriers). The status of the actor, on the other hand, appeared to have a direct relationship to secondary appraisal (i.e., the barriers variable) which was separate from the relationship between appraisal and stressfulness. The relationships were independent of negative affectivity.

Discussion

The purpose of this study was to examine how incidents of workplace violence are experienced by victims, and more specifically, to investigate the similarities and differences in how incidents of physical, sexual, and emotional violence are cognitively appraised.

Nature and Extent of Violence

Based on the checklist and the incidents detailed by participants, emotional incidents were the most frequently described, followed by the sexual and physical incidents, respectively. Further, the physical incidents which were most frequently reported were those which involved threats of physical attack rather than those which involved actual attacks. The sexual incidents most frequently reported were those involving gender harassment (e.g., "been subjected to terms of endearment") as opposed to sexual coercion or unwanted sexual attention. In addition, when participants had examples of emotional, physical, and/or sexual incidents from which to choose they were more likely to chose an emotional incident as most "memorable". Hence these results suggest that not only did the emotional incidents occur more frequently than the physical and sexual incidents in this sample, they were also perceived as more memorable by participants. These findings are consistent with Baron and Neuman's (1996) study of workplace aggression that found that verbal and passive forms of aggression were rated as more frequent by employees than physical and active forms of aggression. Hence, without discounting the importance and tragic nature of acts of physical violence in the workplace, these results challenge the research and media reports which focus solely on physical violence at work.

Which forms of violence co-occurred?

This study is one of few which has examined the three forms of violence concurrently. As a result, it was informative to examine how often incidents of physical, sexual, and emotional violence occurred independently of each other, and how often they co-occurred. Emotional and sexual incidents most often occurred independently of other forms of violence, whereas the majority of the physical incidents involved both physically and emotionally violent behaviors. These results are consistent with findings in the family abuse which has shown that emotional, physical, and sexual forms of violence often co-occur (Crittenden et al., 1994; McLaren & Brown, 1989; Ney, 1987). These findings of co-occurrence indicate the importance of investigating the various forms of violence concurrently, and the inherent limitations in trying to 'tease' these forms apart in order to examine them in isolation. Further, the family abuse research suggests that emotional or psychological abuse often precedes and predicts future physical abuse (Murphy & O'Leary, 1989). Although not investigated in this study, this sequence of emotional, or perhaps sexual violence, leading to physical violence may also occur. This would be an interesting question for future research in this area.

Who were the actors?

One of the advantages of having examined these forms of violence together is that our study revealed that different actors (as defined by organizational position and sex) engaged in different forms of violence. Incidents of emotional violence were most frequently perpetrated by coworkers, followed by clients and supervisors. These results are consistent with other studies in the area of emotional abuse which indicate that actors of emotional violence are often co-workers (e.g., Keashly et al., 1994; Price Spratlen, 1994; Price Spratlen, 1995). Further, many of the studies with health care professionals suggest that clients frequently verbally abuse their caregivers (Goodridge et al., 1996; Graydon et al., 1994). The emotional incidents in this study were perpetrated by female actors more than male actors. This is an interesting finding since several studies in the area of emotional violence have not found sex differences among the actors of this form of violence. One possible explanation is that this study occurred in a female-dominated health care setting. Since the incidents of emotional violence in this study came mostly from coworkers, it is not surprising that most of the actors of these incidents were women.

Consistent with the sexual harassment literature (Betz & Fitzgerald, 1987; Cleveland, 1994; Gutek, 1985; Schneider et al., 1997), most of the sexual incidents in this study were perpetrated by men. Further, these sexual incidents were most frequently perpetrated by clients, followed by coworkers. These results are also consistent with the research which suggests that employees are more likely to be sexually harassed by a coworker than a supervisor (Cleveland, 1994), and with the health care area research which suggests that clients are often the actors of sexual harassment (Cox, 1987; Ford Williams, 1996).

With regards to the incidents of physical violence, these incidents were overwhelmingly perpetrated by clients. This finding is consistent with the literature in the workplace physical violence field which suggests that physical violence almost always comes from someone outside of the organization (NIOSH, 1997). This literature also suggests that the actors of physical violence tend to be male (Bulatao & VandenBos, 1996; Rosenthal et al., 1992). The present study also found that the incidents of physical violence were most often perpetrated by male actors. This finding could be a function of the ratio of male and female clients in the settings sampled, could be related to the tendency for males to be more physically aggressive than females, or could be related to other factors not considered here.

The Dynamics of Context, Appraisal, and Stress .

As was hypothesized, participants differed in their appraisals of the three forms of violence. Both the meaningfulness, or importance of incidents of violence, and the ability to manifest control over these incidents differed across incidents of physical, sexual, and emotional violence. The physical and emotional incidents were perceived as the most meaningful incidents (i.e., central and important), followed by the sexual incidents. This ordering was slightly different from what was expected. The original hypothesis was based on the evidence from an earlier study (Rogers, 1998) that the physical incidents were generally perceived as more severe, or intense, than the sexual and emotional incidents. This hypothesis was partly supported here since the physical incidents were perceived as more meaningful than the sexual incidents. Counter to this hypothesis, the emotional incidents were also perceived as more meaningful than the sexual incidents.

With regards to participants' ability to control incidents of violence, the emotional incidents were perceived as the most difficult with which to manifest control. The sexual incidents were perceived as the next most difficult group of incidents to control, followed by the physical incidents. In other words, the physical incidents were perceived as the easiest incidents with which to manifest control. These findings support the hypothesis that incidents of physical and sexual violence would be the easiest with which to manifest control due to their being proscribed by federal law and organizational policy. In the health care settings sampled in this study, the majority of the incidents of physical and sexual violence reported were the acts of cognitively impaired patients/clients. These institutions recognize the occurrence of these incidents and hence try to provide employees with the resources and support necessary to effectively deal with these types of incidents. No such avenues exist for employees who experience incidents of emotional violence, at least not in most organizations (Diaz & McMillin, 1991; Estrin, 1996; Yamada, 2000).

The variability in appraisal of these three forms of violence were a result of the different contexts within which these behaviors occurred. These results suggest that the experiences of being physically, sexually, or emotionally mistreated cannot be adequately examined and understood in isolation, except perhaps for the most extreme examples of each form (e.g., being sexually assaulted). Several contextual variables were proposed in this study, of which only a subset were related to the appraisal variables. Intent to harm and actor status were demonstrated as important in targets' appraisal. Incidents in which the actor was perceived as intending to harm the victim were appraised as more central and important to the victim, and were associated with lower levels of personal control and with greater barriers to action. In other words, if the recipient felt that the actor wanted to hurt him or her with his or her actions, as opposed to viewing the incident as an accident, the victim was more likely to appraise the incident as a threat.

Although the findings related to intent to harm and primary appraisal were as hypothesized, and are consistent with the research in the areas of spousal satisfaction and sexual harassment (Bradbury & Fincham, 1990; Pryor & Day, 1988), the findings related to secondary appraisal were not hypothesized. A possible explanation for this finding involves considering how incidents in which the victim perceives negative intent may be less predictable in nature. If an individual is treated violently and perceives that the incident was an accident they are likely to believe that the event was isolated, whereas if the individual perceives that the incident was intended they may worry that they cannot respond to the incident. This unpredictability may make acting during the event, and dealing with the event after the fact, more difficult, hence explaining the relationship between negative intent and control.

Although intentions to harm help determine whether an incident will be perceived as stressful, other contextual variables also play a role in these perceptions, resulting in situations in which a actor may not have intended to hurt the target by his/her actions but where the target still experienced the situation as threatening. Status of the actor was related to the importance and centrality of an incident, as well as the personal control and barriers associated with the incident. Those incidents perpetrated by supervisors were associated with the highest level of importance, centrality, and barriers, and with the lowest levels of personal control, followed by incidents perpetrated by coworkers, subordinates, and patients, respectively. In other words, incidents perpetrated by patients were associated with the lowest levels of meaningfulness and the highest levels of control. Hence, as hypothesized, and in accord with the sexual harassment literature (Sheffey & Tindale, 1992), those incidents perpetrated by higher status actors (i.e., supervisors and coworkers) were appraised as more meaningful and more difficult to control than those incidents perpetrated by lower status actors (i.e., subordinates or clients). Several study participants discussed the differences between being mistreated by patients and being mistreated by fellow workers. Participants explained how they felt that their training had provided them with the skills necessary to deal with the violence which comes from patients. Further, participants felt that they had the necessary support from coworkers and from the organization to deal with these incidents. Conversely, participants often felt ill-prepared and ill-supported to deal effectively with incidents of violence from coworkers and supervisors. Participants often mentioned not knowing who to talk to in the organization or how to proceed in order to deal with such incidents. This is consistent with the finding in the emotional abuse area that employees often perceive little support in dealing with coworker abuse (Keashly, 1998).

Although these results suggest that incidents perpetrated by patients were less likely to be appraised as stressful than those perpetrated by fellow workers,it is important not to downplay the impact of abuse by patients. In its most extreme forms, patient aggression can have a serious impact on the physical and psychological well-being of health care workers, and even result in death (Boyd, 1995; Goodridge et al., 1996; Liss & McCaskell, 1994). Fortunately these extreme incidents occur infrequently. What are more frequent are the less severe physical, verbal, and sexual incidents. Many of the participants in this study indicated that they have learned to accept these less severe incidents of violence as "part of the job". However, the workplace violence research which examines health care professionals indicates that over time, day-after-day, these abusive incidents can take a significant toll (Cox, 1990; Diaz et al., 1991; Lanza, 1983). A quote from one participant summarizes this phenomenon: "...dealing successfully with out-of-control clients still remains a situation which can cause personal stress...we deal with them all the time, but they are not run-of-the-mill".

Three other contextual variables were hypothesized to be related to the appraisal variables in this study. These hypotheses were not supported by the data. First, the sex of the actor was originally proposed as a characteristic of the actor which may influence incident appraisals. Although there were sex differences among the actors who perpetrated the three forms of violence (i.e., males were more likely to perpetrate incidents of physical and sexual violence and females were more likely to perpetrate incidents of emotional violence), these sex differences did not explain unique variance in the appraisal variables. This finding is counter to the sexual harassment literature which suggests that incidents perpetuated by males are generally perceived as more abusive (Fitzgerald & Shullman, 1993; Gutek, 1985). The results of the present study suggest that across incidents of workplace violence, including physical, sexual, and emotional incidents, the sex of the actor was not a salient factor in stress appraisals. What appeared more important than the sex of the actor was his/her job status, or position relative to the target. This finding is consistent with research which has been done in the area of organizational influence and negotiation. When examining how individuals use negotiating and bargaining behaviour in organizations researchers in this area have found that situational power is a better predictor of negotiator behaviour than is sex (Kanter, 1977; Watson, 1994).

Second, history with the actor was not identified as a salient contextual feature in our analyses. This finding may have more to do with the inclusion of intent to harm as a variable in the same analyses. In the marital satisfaction literature (e.g., Bradbury & Fincham, 1990), causes of negative events are generally seen as internal to the partner and that one of the internal attributions is intent. Thus, to the extent history of actor (consistency in Kelley's (1973) terminology) affects causal attributions, it will affect intent to harm. In some earlier work on emotional abuse at work, Keashly, Welstead and Delaney (1996) found that actor history was related to observers' assessment of a behavior as abusive and that this consistency had its effect through its relationship to intent to harm. Following this logic, actory history may well be important as suggested by workplace abuse researchers (e.g., Hoel et al, 1999; Keashly, 1998) but through its relationship to attributions of intent.

Third, it was originally hypothesized that the severity of an incident would be related to how it is appraised. However, incident severity was not identified in the analyses as a salient contextual feature. Even though severity was correlated with certain appraisal variables, this variable was not related to these dependent variables in the multivariate analyses. This suggests that severity may share variance with the other contextual variables, and may not predict unique variance in cognitive appraisal.

It was originally hypothesized that both primary and secondary appraisal would predict the perceived stressfulness of an incident. This hypothesis was supported, although not all five appraisal variables proposed explained a unique amount of variance in stressfulness. The results suggest that appraisals of importance, centrality and barriers predicted unique variance in perceived stressfulness, while appraisals of personal control and the control by others did not. Although these findings support the original hypothesis, they raise the question of why personal control and the control by others did not aid in this prediction. There are at least two possible reasons for this finding. Other studies investigating the relationships between primary appraisal, secondary appraisal and stressfulness have also had difficulty establishing a direct or moderating effect of control on stressfulness (Dobson & Neufeld, 1979; Folkman et al., 1986a; Thornton, 1992). At the same time, other researchers have had success examining the relationships between secondary appraisal and coping (Folkman et al., 1986b). For example, Folkman et al. (1986b) found that when an incident was perceived as "changeable," individuals used confrontative and planful problem-solving coping styles, while they used distancing and escape/avoidance coping styles when they perceived an incident as "one which needed to be accepted". These findings suggest that appraisals of control are important in the stress process but that these appraisals may play a more important role in predicting how individuals cope with stressful incidents.

Another possible explanation for why perceptions of control were not related to perceptions of stressfulness involves the concept of barriers as used in this study. The barriers measure assessed perceived obstacles internal to the organization which prevented individuals from responding to incidents in the way they would have preferred. The control-self and control-others scales, on the other hand, assessed whether individuals possessed, or had access to the resources necessary to deal with incidents. It is possible that even if an individual possessed, or had access to these resources, they could not act upon them because of the presence of certain barriers. Hence, the concept of barriers to responding may be an important moderator in the appraisal-stress relationship and needs to be considered in any discussion of secondary appraisal.

It was hypothesized that the relationship between the contextual variables and perceived stressfulness would be mediated by the cognitive appraisal variables. With regards to intent to harm, this original hypothesis was partly supported. That is, the results suggest that the relationship between intent to harm and stressfulness was partially mediated through the primary appraisal variables (i.e., importance and centrality) and was partially direct. One of the reasons intent to harm was uniquely related to both appraisal and stressfulness may lie in its conceptualization as a contextual variable. This variable involved the assessment, or appraisal, of intent. It is possible, therefore, that intent to harm resembles an appraisal variable more than a contextual variable. Tying this together with the argument for the findings around history of the actor, history may well be the contextual variable and intent is an important aspect of the appraisal process.

In the case of the relationship between actor status and stressfulness, the hypothesis was not supported. The tests for mediation suggest that status of the actor was related to the secondary appraisal variables but not to perceived stressfulness. Hence the appraisal variables could not have possibly mediated this relationship. That is, the relationship between actor status and appraisal was separate from the relationship between appraisal and stressfulness. Unlike intent to harm, actor status was only related to secondary appraisal, and hence was only related to part of the appraisal construct. The results suggest that both primary and secondary appraisal were necessary in the prediction of stressfulness. Hence actor status may not have had a large enough influence on the cognitive appraisal variables as a whole to have a significant impact on stressfulness.

Summary

The results of this study shed some light on how people subjectively experience incidents of workplace violence, and those variables which are important in this experience. To begin, these results indicate that participants in this study appraised incidents of physical, sexual, and emotional violence differently. However, this relationship between form of violence and appraisal was explained, and statistically disappeared, when certain contextual variables were taken into consideration. The contextual variables which were identified as being particularly salient in appraisals were intent to harm and actor status. These findings suggest that participants differed in their evaluations of physical, sexual, and emotional violence not because of inherent differences between these incidents, but because of the context in which the incidents occurred. Hence by only considering the form of violence involved in an incident, and by not considering other relevant contextual variables, researchers may portray a partial, or perhaps even distorted picture of what gets experienced as workplace violence.

The results of this study also suggest that both primary and secondary appraisal variables predicted stressfulness, but that appraisals of importance, centrality, and barriers were particularly useful in this prediction. That is, the more important and central the incident, and the greater barriers associated with the incident, the more stressful the experience. It is also important to note that the relationships which were identified in this study were found independent of the influences of trait negative affectivity.

Limitations

The limitations of any study methodology are intimately connected to the purported research questions. It can be argued that the self-report and retrospective nature of this study is inherently problematic in terms of whether the reports provided are accurate (i.e., confirmed by observers). The self-report nature is actually a key strength for this study which focused on people's perceptions and cognitions about their experiences with violence (Brodsky, 1976). However, sole reliance on self-report data raises the specter that the relationships examined may be inflated due to mono-source error. By demonstrating that the hypothesized relationships were not a function of trait negative affectivity, we have challenged this potential limitation of our results.

Regarding the retrospective nature of the data, attempts were made to minimize these distortions by limiting the amount of time between the incident and the assessment of the incident. The time period between the occurrence of the incident and its assessment ranged from five to thirty days in this study. Given that many studies in this area ask respondents to recall and discuss incidents from six months to a year ago, the likelihood of these types of distortions were less in this study than in many (Hoel et al, 1999; Keashly, 1998).

The cross-sectional nature of this study is a potential limitation because causal connections among the variables can not be conclusively established. However, our intent for this study was to provide much needed data on the nature, extent, and experience of workplace violence in its various forms. To meet this goal, we felt it was more appropriate to look at a large number and variety of incidents at one point in time instead of a few incidents over several points in time. Thus, these results and our proposed interpretations opens the door for future longitudinal research needed to establish causality.

Finally, health care institutions were sampled in the present study. This sampling strategy is appropriate given the nature of our research questions. However, it may limit generalizability to other settings. For example, the fact that actor status was so salient in the appraisals of incidents in this study indicate that some aspects of the findings may be health care related. The relationship between a health care professional and a client is a rather unique one. It would be interesting to investigate whether actor status is a salient contextual feature in other client-service provider relationships and with workers in other industries. However, Lazarus and Folkman (1984) argue that the cognitive appraisal process through which potentially stressful incidents are appraised does not differ based on setting. Thus, the relationships between the contextual variables and the appraisal variables found in this study should generalize across settings. What may differ across settings are the specific contextual features which are salient in predicting appraisal and the actual levels of primary and secondary appraisal.

Implications for Research

As one of the few studies to empirically examine physical, sexual, and emotional forms of violence concurrently, this study contributes to the area of workplace violence in numerous ways. First, the results suggest the relative frequency with which incidents of workplace violence occur in health care settings. The fact that only 22 participants (9%) in this study returned blank checklists provides some indication of the incidence and seriousness of this issue. Further, the results provide the area of workplace violence with information on the relative frequency with which the different forms of violence occur in health care settings, as well as an idea of who generally perpetrates the three forms. Moreover, this study provides some initial evidence that certain forms of violence co-occur. The child abuse literature suggests that the experience of multiple forms of abuse is associated with greater physical and psychological stress than the experience of a single form of violence reactions (Moeller, Bachmann, & Moeller, 1993). Future research in the area of workplace violence might examine whether the experience of more than one form of violence is associated with greater, or more intense stress reactions than one form of violence on its own.

Second, this study points to the importance of context in understanding the experience of workplace violence. The contextual nature of sexual harassment has been established in the literature (York, 1989; Pryor & Day, 1988). This study suggests that physical and emotional forms of violence are also contextual in nature. This is an instructive finding since many people assume that the experience of physical violence is less dependent, if dependent at all, on context. Two contextual features, intent to harm and actor status, were particularly helpful in explaining the relationship between form of violence and appraisal, and the experience of workplace violence more generally. Researchers need to identify other salient contextual features which play a role in appraisals of workplace violence (e.g., predictability and distinctiveness of incident).

Last, the results of this study suggest that primary and secondary appraisal predict incident stressfulness in the context of workplace violence. However, neither the control-self nor the control-others scales added to this prediction. Future research needs to examine the relationship between these two control-related variables and the barriers variable in order to determine whether the control-related variables add to the prediction of stressfulness in other contexts. These studies should also examine the process through which people cope with incidents of workplace violence and the role that control plays in the selection of coping strategies (Keashly, 1998). Lazarus and Folkman's (1984) stress appraisal framework appears to be a useful framework through which to understand and examine the experience of workplace violence. Hence we would suggest that this framework be employed in these future research studies.

Implications for Practice

The first implication of this study relates to how practitioners define and view workplace violence. Many organizations do not recognize the existence nor incidence of emotional forms of workplace violence (Diaz & McMillin, 1991; Estrin, 1996). This study suggests that emotional violence in health care settings is not only the most frequently experienced form of violence but it is also among the most meaningful and the most difficult to control. Further, practitioners need to understand that the three forms do not occur in isolation. For this reason organizations need to develop policies and guidelines which address the full spectrum of workplace violence. Many organizations have policies which address sexual and physical forms of violence. Not only should these policies be modified to include emotional violence, these policies should reflect the integrated nature of these three forms of violence.

Second, organizations must recognize that the experience of workplace violence is contextual in nature. In other words, the exact same behaviour may be appraised as stressful by one person in one context, and may be appraised as neutral by another person in another context. For this reason, organizations cannot address workplace violence by just generating and posting a list of behaviours which are prohibited. Organizations could never capture all of the potentially threatening behaviours and situations. A more effective strategy for addressing workplace violence is to develop policies and guidelines which include a strong statement of the organization's philosophy and commitment concerning workplace violence, which discuss what workplace violence involves, and which outline how the organization expects its employees to treat one another.

The role of organizational norms in the occurrence of workplace violence has been suggested by sexual harassment researchers who propose that workplace norms and organizational climate can inhibit sexual harassment (Fitzgerald, et al., 1997; Pryor & Day, 1988). Specifically, it has been suggested that sexual harassment is the result of an individual predisposition to harass, and an organizational climate that allows or supports such behaviour (Fitzgerald, et al., 1997). Hence for an organization to effectively address the issue of violence in their workplace they must model and reward appropriate behavior, as well as consistently punish inappropriate behavior even in its mildest forms.

Another practical implication concerns perceived barriers in the organization. The results of this study suggest that if an individual perceives that there are barriers internal to the organization which prevent him/her from dealing with an incident, he/she will be more likely to perceive the incident as stressful. Fitzgerald et al. (1997) highlight how barriers can influence the occurrence of sexual harassment. Specifically, these researchers indicate that those employees who believe that their organization is tolerant of sexual harassment experience higher levels of harassment. Based on these findings in the sexual harassment area and the findings of the present study, it would appear that perceptions of organizational tolerance and organizational barriers influence both the occurrence and experience of workplace violence. Indeed, a recent study of targets of emotionally abusive behaviors found that organizational responding was a key factor in the extent to which the respondents felt they had been treated abusively (Keashly, in press). Together, these findings relate to perceptions of interpersonal justice, or the degree to which an organization demonstrates sensitivity and concern for its employees (Greenberg, 1993). Employees want to feel that the organization genuinely cares about them and that the organization treats all of its employees fairly. Hence, if organizations cannot prevent incidents of workplace violence, they must at least identify and remove the barriers which are preventing their employees from reporting and dealing with these incidents.

References

Arnetz, J.E. (1998). The Violent Incident Form (VIF): A practical instrument for the registration of violent incidents in the health care workplace.Work & Stress, 12(1), 17-28.

Baldwin, D.C., Daugherty, S.R., & Eckenfels, E.J. (1991). Student perceptions of mistreatment and harassment during medical school: A survey of ten United States schools. The Western Journal of Medicine, 155, 140-145.

Barling, J., Dekker, I., Loughlin, C.A., Kelloway, E.K., & Fullager, C. (1993). Prediction and replication of the organizational and personal consequences of workplace seexual harassment. Working Paper, Queen’s University, Kingston, Ontario.

Baron, R.A. & Neuman, J.H. (1996). Workplace violence and workplace aggression: Evidence on their relative frequency and potential causes. Aggressive Behavior, 22, 161-173.

Baron, R.M. & Kenny, D.A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statstical considerations. Journal of Personality and Social Psychology, 51(6), 1173-1182.

Betz, N.E. & Fitzgerald, L.F. (1987). The career psychology of women. San Diego, CA: Academic Press.

Boyd, N. (1995).Violence in the workplace in British Columbia: A preliminary investigation. Canadian Journal of Criminology, 37(4), 491-519.

Bollen, K., & Lennoz, R. (1991). Conventional wisdom on measurement: A

structural equations perspective. Psychological Bulletin, 110, 305-314.

Bradbury, T.N. & Fincham, F.D. (1990). Attributions in marriage: Review and critique. Psychological Bulletin, 107(1), 3-33.

Brief, A.P., Burke, M.J., George, J.M., Robinson, B.S., & Webster, J. (1988). Should negative affectivity remain an unmeasured variable in the study of job stress? Journal of Applied Psychology, 73(2), 193-198.

Bulatao, E.Q., & VandenBos, G.R. (1996). Workplace violence: Its scope and the issues. In G.R. VandenBos and E.Q. Bulatao (Eds.) Violence on the job: Identifying risks and developing solutions. Washington, DC: American Psychological Association.

Chappell, D. & Di Martino V. (1998). Violence at work. Geneva: International Labor Office.

Cleveland, J.N. (1994). Women and sexual harassment: Work and well-being in US organizations. In M.J. Davidson & R.J. Burke (eds). Women in management: Current research issues. London: PCP Publishing Ltd.

Cohen, J. (1960). A coefficient of agreement for nominal scales: Educational and Psychological Measurement, 10, 37-46.

Cox, H. (1987). Verbal abuse in nursing: Report of a study. Nursing Management, 19, 47-50.

Cox, H. (1990). Verbal abuse nationwide, part II: Impact and modifications. Nursing Management, 22, 66-69.

Crittenden, P.M., Claussen, A.H. & Sugarman, D.B. (1994). Physical and psychological maltreatment in middle childhood and adolescence. Development and Psychopathology, 6, 145-164.

Dewe, P. (1991). Primary appraisal, secondary appraisal and coping: Their role in stressful work encounters. Journal of Occupational Psychology, 64 331-351.

Dewe, P. (1992). Applying the concept of appraisal to work stressors: Some exploratory analysis. Human Relations, 45(2), 143-164.

Diaz, A.L., & McMillin, J.D. (1991). A defiition and description of nurse abuse. Western Journal of Nursing Research, 13(1), 97-109.

Dobson, K.S. & Neufeld, W.J.(1979). Stress-related appraisals: A regression analysis. Canadian Journal of Behavioral Science, 11, 274-285.

Einarsen, S. (1999). The nature and causes of bullying at work. International Journal of Manpower, 20(1/2), 16-27.

Elliott, T.R., Chartrand, J.M., & Harkins, S.W. (1994). Negative affectivity, emotional distress, and athe cognitive appraisal of occupational stress. Journal of Vocational Behavior, 45, 185-201.

Estrin, C.B. (1996). Emotional abuse in the workplace: Breaking the code of silence. Legal Assistant Today, 78-79.

Fitzgerald, L.F., Drasgow, F., Hulin, C.L., Gelfand, M.J., & Magley, V.J. (1997). Antecedents and consequences of sexual harassment in organizations: A test of an integrated model. Journal of Applied Psychology, 82(4), 578-589.

Fitzgerald, L.F. & Schullman, S.L. (1993). Sexual harassment: A research analysis and agenda for the 1990s. Journal of Vocational Behavior, 42, 97-109.

Folkman, S. & Lazarus, R.S. (1984). If it changes, it must be process: Study of emotional and coping during three stages of a college examination. Journal of Personality and Social Psychology, 48(1), 150-170.

Folkman, S., Lazarus, R.S., Gruen, R.J. & DeLongis, A. (1986a). Appraisal, coping, health status, and psychologocial symptoms. Journal of Personality and Social Psychology, 50(3), 571-579.

Folkman, S., Lazarus, R.S., Dunkel-Schetter, C., DeLongis, A. & Gruen, R.J. (1986b). Dynamics of a stressful encounter: Cognitive appraisal, coping, and encounter outcomes. Journal of Personality and Social Psychology, 50(5), 992-1003.

Ford Williams, M.F. (1996). Violence and sexual harassment: Impact on registered nurses in the workplace. American Association of Occupational Health Nurses Journal, 44(2), 73-77.

Goodridge, D.M., Johston, P., & Thomson, M. (1996). Conflict and aggression as stressors in the work environment of nursing assistants: Implications for institutional elder abuse. Journal of Elder Abuse & Neglect, 8(1), 49-67.

Graydon, J., Kasta, W. & Khan, P. (1994). Verbal and physical abuse of nurses. Canadian Journal of Nursing Administration, 7(4), 70-89.

Greenberg, J. (1993). The social side of fairness: Interpersonal and informational classes of justice. In R. Cropanzano (Ed.) Justice in the workplace: Approaching fairness in human resource management. Hillsdale, NJ: Lawrence Erlbaum.

Gutek, B.A. (1985). Sex and the workplace. San Francisco: Jossey-Bass Inc.

Hoel, H., Rayner, C., & Cooper, C. L. (1999). Workplace bullying. In C.L. Cooper & I.T. Robertson (Eds). International Review of Industrial and Organizational Psychology, 14, 195-230.

Kahn, R.L. & Byosiere, P. (1992). Stress in organizations. In M.D. Dunnette & L.M. Hough (Eds.), Handbook of industrial and organizational psychology, (Vol 3) Pp. 242-298. Palo Alto, CA: Consulting Psychologists Press.

Keashly, L (in press). Interpersonal and systemic aspects of emotional

abuse at work: The target’s perspective. Violence and Victims, 20.

Keashly, L. (1998). Emotional abuse in the workplace: Conceptual and empirical

issues. Journal of Emotional Abuse, 1(1), 85-115.

Keashly, L, Trott, V., & MacLean, L.M. (1994). Abusive behavior in the

workplace: A preliminary investigation. Violence and Victims, 9(4), 125- 141.

Keashly, L., Welstead, S. & Delaney, C. (1996). Perceptions of abusive behaviors in the workplace: Role of history, emotional impact, and intent. Unpublished manuscript, University of Guelph, Ontario.

Kelley, H.H., (1973). The processes of causal attribution. American Psychologist, 28, 107-128.

Klecka, W.R. (1980). Discriminant analysis. Beverley Hills, CA: Sage.

Lanza, M.L. (1983). The reactions of nursing staff to physical assault by a patient. Hospital and Community Psychiatry, 34, 44-47.

Liss, G.M. & McCaskell, L. (1994). Injuries due to violence: Worker’s compensation claims among nurses in Ontario. American Association of Occupational Health Nurses Journal, 42(8), 384-390.

McLaren, J. & Brown, R.E. (1989, Sept.) Childhood problems associated with abuse and neglect. Canada’s Mental Health, 1-6.

Moeller, T.P, Bachmann, G.A., & Moeller, J.R. (1993). The combined effects of physical, sexual, and emotional abuse during childhood: Long-term health consequences for women. Child Abuse & Neglect, 17, 623-640.

Murphy, C.M. & O’Leary, K.D. (1989). Psychological aggression predicts physical aggression in early marriage. Journal of Consulting and Clinical Psychology, 57, 579-582.

National Institute for Occupational Safety and Health (1997, June). Violence in the workplace (press release). [on-line]. Available: http://www.cdc.gov/niosh/violfs.html.

Neuman, J.H. & Baron, R.A. (1997). Aggression in the workplace. In R.A. Giacalone & J. Greenberg (Eds.) Antisocial behaviors in organizations. Chpt 3, pp. 37-67. Thousand Oaks, CA: Sage.

Ney, P.G. (1987). Does verbal abuse leave deeper scars: A study of children and parents. Canadian Journal of Psychiatry, 32(5), 371-378.

Northwestern National Life Insurance Company (1993). Fear and violence in the workplace. Minneapolis: Northwestern National Life Insurance Company.

O'Leary-Kelly, A.M., Griffin, R.W., & Glew, D.J. (1996). Organization- motivated aggression: A research framework. Academy of Management Review, 21(1), 225-253.

Paterson, R.J. & Neufeld, R.W.J. (1989). The stress response and parameters of stressful situations. In R.W.J. Neufeld (Ed.) Advances in the investigation of psychological stress. New York: Wiley.

Peacock, E.J. & Wong, P.T. (1990). The stress appraisal measure (SAM): A multidimensional approach to cognitive appraisla. Stress Medicine, 6, 227-236.

Price Spratlen, L. (1994). Perceived workplace msitreatment in higher education: Characteristics and consequences. American Association of Occupational Health Nurses Journal, 42(11), 548-554.

Price Spratlen, L. (1995). Interpersonal conflict which includes mistreatment in a university workplace. Violence and Victims, 10(4), 285-297.

Pryor, J.B. (1985). The lay person’s understanding of sexual harassment. Sex Roles, 13(5/6), 273-286.

Pryor, J.B. & Day, J.D. (1988). Interpretations of sexual harassment: An attributional analysis. Sex Roles, 18(7/8), 405-417.

Richman, J.A., Rospenda, K.M., Nawyn, S.J., Flaherty, J.A., Fendrich, M., Drum, M.L., & Johnson, T.P. (1999). Sexual harassment and generalized workplace abuse among university employees: Prevalence and mental health correlates. American Journal of Public Health, 89(3), 358-363.

Robinson, S.L. & Bennett, R.M. (1995). A typology of deviant workplace behaviors: A multidimensional scaling study. Academy of Management Journal, 38(2), 555-572.

Rogers, K. (1998). Toward an integrative understanding of workplace mistreatment. Unpublished Doctoral dissertation, University of Guelph, Guelph, Ontario.

Rosenthal, R.L., Edwards, N.B., Rosenthal, R.H., & Ackerman, B.J. (1992). Hospital violence: Site, severity, and nurses’ preventive training. Issues in Mental Health Nursing, 13(4), 349-356.

Rosen, H.M. & Korabik, K. (1991). Workplace variables, affective responses and intention to leave among women managers. Journal of Occupational Psychology, 64(4), 317-330.

Sackett, P.R. & Larson, J.R. (1990). Research strategies and tactics in industrial and organizational psychology. In M.D. Dunnette and L.M. Hough (Eds.) Handbook of industrial and organizational psychology. Palo Alto, CA: Consulting Psychologists Press.

Schneider, K.T., Swan, S. & Fitzgerald, L.F. (1997). Job-related and psychological effects of sexual harassment in the workplace: Empirical evidence from two organizations. Journal of Applied Psychology, 82(3), 401-415.

Sheffey, D. & Tindale, R.S. (1992). Perceptions of sexual harassment in the workplace. Journal of Applied Social Psychology, 22(19), 1502-1520.

Silver, H.K. & Glicken, A.D. (1990) Medical student abuse: Incidence, severity, and significance. The Journal of the American Medical Association, 263(4), 527-532.

Solomon, Z., Mikulincer, N. & Benbenishty, R. (1989). Locus of control and combat-related post-traumatic stress disorder: The intervening role of battle intensity, threat appraisal, and coping. British Journal of Clinical Psychology, 28, 131-144.

Thornton, P.I. (1992). The relation of coping, appraisal, and burnout in mental health workers. The Journal of Psychology, 126(3), 261-271.

Vinokur, A.D., Threatt, B.A., Vinokur-Caplan, D., & Satariano, W.A. (1990). The process of recoery from breast cancer in younger and older patients. Cancer, 65, 1242-1254.

Vitaliano, P.P., Russo,J., Weber, L., & Celum, C. (1993). The dimensions of stress scale: Psychometric properties. Journal of Applied Psychology, 23, 1847-1878.

Wanous, J.P. & Reichers, A.E. (1996). Estimating ther reliability of a single item measure. Psychological Reports, 78, 631-634.

Watson, D., Clark, L.A. & Tellegen, A. (1988). Devlopment and validation of brief measures of positve and negative affect. The PANAS scales. Journal of Personality and Social Psychology, 54, 1060-1070.

Whittington, R., Shuttleworth, S., & Hill, L. (1996). Violence to staff in a general hospital setting. Journal of Advanced Nursing, 24, 326-333.

Yamada, D. (2000). The phenomenon of "workplace bullying" and the need for status-blind hostile work environment protection. Georgetown Law Review.

York, K.M. (1989). Defining sexual harassment in workplaces: A policy capturing approach. Academy of Management Journal, 32(4), 830-850.