Journal of Affective Disorders 88 (2005) 131 136

www.elsevier.com/locate/jad

Research report

Variability in suicidal ideation: A better predictor of suicide

attempts than intensity or duration of ideation?
Tracy K. Witte a, Kathleen K. Fitzpatrick b, Thomas E. Joiner Jr. a,
Norman Bradley Schmidt a,*
a

Department of Psychology, Florida State University, Tallahassee, FL 32306-1270, United States

b
The Ohio State University, United States
Received 4 April 2005; received in revised form 17 May 2005; accepted 31 May 2005
Available online 27 July 2005

Abstract
Background: Much of the suicidology literature focuses on establishing contextual risk factors for suicidal behavior. However,
the study of the parameters of suicidal behavior (e.g., intensity, duration, and variability) has been somewhat neglected [Joiner,
T., Rudd, M.D., 2000. Intensity and duration of suicidal crises vary as a function of previous suicide attempts and negative life
events. J. Consult. Clin. Psychol. 68, 909916]. Having previously established a relationship between variability in suicidal
ideation and a previous history of suicide attempts [Witte, T.K., Fitzpatrick, K.K., Warren, K.L, Schatschneider, C., Schmidt,
N.B., submitted for publication. Naturalistic Evaluation of Suicidal Ideation: Variability and Relation to Attempt Status], we felt
it important to assess the liability conferred by a variable pattern of ideation compared to the intensity and duration of suicidal
thoughts. We also examined if there was an interaction between gender and the parameters of intensity, duration, and variability.
Method: One hundred eight participants (54 non-attempters, 35 single attempters, and 19 multiple attempters) completed the
Suicide Probability Scale every day for 4 weeks, allowing us to measure the parameters of interest. These variables were entered
into a regression model as predictors of previous suicide attempts.
Results: Consistent with prediction, high variability of ideation was the only significant predictor of previous attempt status. In
addition, an interaction between gender and variability in suicidal ideation suggested that variability appeared more critical in
predicting previous attempts for males.
Limitations: The limited number of multiple attempters in our sample and the use of college students limit the current study.
Conclusions: Variability appears to be the most potent predictor of attempt status among the parameters of suicidal ideation
examined in the current study. This relationship appears to be particularly important in males, suggesting that fluctuating levels
of suicidal ideation may confer future risk for suicide.
D 2005 Published by Elsevier B.V.
Keywords: Suicide; Variability; Ideation

* Corresponding author. Tel.: +1 850 644 1707; fax: +1 850 644 7739.
E-mail address: Schmidt@psy.fsu.edu (N.B. Schmidt).
0165-0327/$ - see front matter D 2005 Published by Elsevier B.V.
doi:10.1016/j.jad.2005.05.019

132

T.K. Witte et al. / Journal of Affective Disorders 88 (2005) 131136

Considerable research has focused on attempting to

better understand the risk factors of suicide. Although
this is clearly an important task, reframing our focus
toward elucidating the parameters of suicidal behavior, such as the duration and intensity of suicidal
ideation, represents an important and relatively unstudied area. The intensity of suicidal ideation, which
was conceptualized by Joiner and Rudd (2000) as the
severity of suicidal symptoms, has a relatively
obvious association with risk factors. It seems logical
that thoughts about suicide could be the beginning of
a trajectory toward suicide completion for some people. Intensity of ideation has been associated with a
history of previous suicide attempts (Joiner et al.,
2000), future suicide attempts (Lewinsohn et al.,
1996), and future suicide completion (Beck et al.,
1999).
Sustained levels of ideation have been posited to
create bwearQ on an individual by depleting their
resources and allowing for an increase in resolve
and commitment to a suicidal act (Joiner, in press).
Consistent with this idea, the duration of suicidal
crises has also been related to suicide attempts (Joiner
and Rudd, 2000; Neuringer, 1982). However, this
relationship appears to be moderated by the presence
of negative life events. For example, Joiner and Rudd
(2000) found that the duration of a suicidal crisis was
considerably longer only for multiple suicide attempters experiencing negative life events.
Variability in suicidal ideation not only likely captures elements of both intensity and duration but also
likely contains elements of affective lability (Harvey
et al., 1989). Affective lability might also contribute to
feelings of a lack of control, which has been associated with thoughts about death and suicidal ideation
(Orbach et al., 2001). In fact, there is evidence that
affectively unstable individuals utilize suicidal behavior as a means of controlling their fluctuating emotions (Zlotnick et al., 1997).
Sex differences in suicidal behavior are important
to consider, including their relation to the parameters
of variability, intensity, and duration. Sex differences
in phenomena related to overt suicidal behaviors
have been relatively well-established (Andrews and
Lewinsohn, 1992; Townsend et al., 2001). However,
no study has yet examined the interaction between
gender differences on the one hand, and the parameters of intensity, duration, and variability of suici-

dal ideation on the other, in relation to suicide

attempts.
The primary aim of the current study was to
examine the three ideation parameters described
above (i.e., intensity, duration, and variability) in
order to determine their unique relations to suicide
attempt status. A unique aspect of the current report
was the utilization of daily, longitudinal data collection to provide a more accurate picture of the nature
and changes in ideation. The second study aim was
to explore the relationship between these parameters
and gender, due to the prominent differences in
pathways to suicidal behavior between the sexes.
Our predictions were as follows. First, we predicted
that variability, because it contains elements of intensity, duration, and affective lability, would be the
best predictor of prior suicide attempts. We also
predicted an interaction between gender and variability based on Winkler et al.s (2004) recent demonstration that depressed females experience higher
levels of affective lability, although we make this
prediction with caution given that Winkler and colleagues studied depressed inpatients, whereas our
sample consists of college students.

1. Method
1.1. Participants
Participants (60% female; mean age = 19.3,
SD = 3.06) were selected from introductory psychology classes and received course credit for their participation. Individuals were selected for participation
following a screening with the Beck Suicide Scale
(BSS; Beck and Steer, 1993), which includes an
item assessing past suicide attempts. Participant selection occurred in two waves. In the first wave, all
participants who responded to screening criteria
were enrolled, regardless of attempt history. The second wave of participant selection focused on targeted
recruitment of those endorsing one or more suicide
attempts. As such, participants with past suicide
attempts were oversampled from the subject pool.
Attempt status was later confirmed during a structured
clinical interview, which includes a suicide history and
risk assessment. Of those screened, 133 entered the
study, but only 108 had complete data and were

T.K. Witte et al. / Journal of Affective Disorders 88 (2005) 131136

used in the final data analyses. Excluded participants

did not differ from included participants in terms of
baseline Beck Suicide Scale (BSS; Beck and Steer,
1993) scores ( F[1, 106] = 1.71, p N .10) or by attempt
status (v 2[2, N = 108] = .883, p N .10). Participants with
past suicide attempts were oversampled from the participant pool.
Participants with at least four consecutive weeks of
data and who were not missing more than two individual data points within the 4-week span were
included in the analyses. If a participant was missing
a data point, we averaged the previous and subsequent
day and utilized the result to replace the missing day.
Individuals were divided into groups based on previous number of suicide attemptsthose with zero
previous attempts (n = 54), one previous attempt
(n = 35), and two or more previous attempts (n = 19).
Gender did not differ across attempt status groups
(v 2[2, N = 108] = 1.96, p N .30) although there was a
slightly higher proportion of females in the single
attempter and multiple attempter groups.
1.2. Measures
1.2.1. Structured clinical interview for diagnosis for
axis I DSM-IV (SCID-IV)
Current and past DSM-IV diagnoses for Axis I
disorders (American Psychiatric Association [APA],
1994) were established through a structured diagnostic interview (First et al., 1994). A graduate student or
undergraduate research assistant trained to criterion
levels of reliability administered the SCID. Training
consisted of reviewing SCID training videos, observations of live interviews, supervised administration and
achievement of 98% inter-rater reliability for diagnosis. All interviews were discussed in conference with
either the second or fourth authors and disagreements
were resolved. Previous attempt status was confirmed
as part of the structured interview, and there were no
disagreements among raters for the attempt status
designation.
1.2.2. Beck Suicide Scale (BSS)
The BSS is a 20-item self-report measure that
reflects the presence of suicide intention, ranging
from suicide wishes to attitudes and plans (Beck and
Steer, 1993). Content, construct, concurrent, and discriminant validity assessments have all shown mod-

133

erate to superior psychometric properties (Beck and

Steer, 1993).
1.2.3. Suicide Probability Scale (SPS)
The SPS is a 36-item, self-report measure of attitudes and behaviors related to suicide risk including a
subscale that measures suicidal ideation specifically
(Cull and Gill, 1995). The SPS good psychometric
properties in both community samples and among
those with prior suicide attempts (Cull and Gill,
1995).
1.3. Analytic procedure
The Mean Squared Successive Difference (MSSD)
statistic was utilized to assess variability in suicidal
ideation (Woyshville et al., 1999). We chose to use
the MSSD because latent growth curve models,
which are generally used to assess individual trajectories as well as group patterns, assume homogenous
and normally distributed curves across individuals.
In the current study, since we were interested in average variability over time, we opted for the MSSD
statistic as a more parsimonious analytic tool. The
formula for this statistical procedure composed of n
elements, the ith element of which is denoted x i , is
given by:
n1
X

MSSD

xi1  xi 2

i1

n1

In a standard linear analysis, variability would be

computed as error, or deviation from the mean or
regression line, whereas the MSSD quantifies the
variation in a single number that can be used as a
variable in regression analysis.
We operationalized intensity, duration, and variability as follows. Intensity was represented by the
overall mean score on the SPS. Duration was conceptualized as the amount of time that the participant
experienced intense ideation and was represented by
the number of days an individual participant had a Tscore of 70 or above on the SPS. This corresponds to
two standard deviations above the mean and is indicative of high levels of suicidal distress. Variability of
suicidal ideation was assessed with the MSSD statistic

134

T.K. Witte et al. / Journal of Affective Disorders 88 (2005) 131136

discussed above. Attempt status was a trichotomous

variable, with a score of 0 indicating no previous
suicide attempts, 1 indicating one previous attempt,
and 2 indicating two or more previous attempts.
1.4. Procedure
Following selection, participants completed a baseline assessment. Those choosing to participate were
administered a structured clinical interview (SCID) to
allow for diagnostic assessment and further evaluation
of suicide risk. Following the interview, participants
completed a packet of self-report measures. Although
we only analyzed data for 4 weeks of questionnaires,
participants were provided six packets containing the
study questionnaires, corresponding to one packet for
each week of study enrollment. Participants were
instructed to complete measures each day for 6
weeks, preferably at the same time each day. Instructions were given both orally and in writing in each
packet. These directions reminded participants to complete the questionnaires based on their feelings and
levels of suicidal ideation that day. To help ensure
compliance, participants returned packets to the
research office weekly. All questionnaires were
marked with an ID number rather than the participants
name in order to ensure confidentiality. All identifying
information, including consent forms, was kept in a
locked office. This study was approved by the Internal
Review Board at The Ohio State University.
1.5. Psychopathology of the participants
Consistent with a college sample partially selected
for a history of suicide attempts, participants showed
some elevations in psychopathology relative to an
unselected college sample. Approximately half (49%)
of the sample met DSM-IV criteria for an Axis I
diagnosis (e.g., 20% Major Depression, 5% Social
Phobia, 4% Post Traumatic Stress Disorder [PTSD],
2% Bipolar I Disorder, 2% Bipolar II Disorder, 2%
past Major Depressive Episode). 15% of non-attempters, 34% of single attempters, and 47% of multiple
attempters met criteria for a current mood disorder
(i.e., Major Depressive Disorder, Dysthymia, Bipolar
I Disorder, Bipolar II Disorder), and these differences
were statistically significant (v 2[2, N = 108] = 9.04,
p b .05).

1.6. Intercorrelations between parameters of

suicidality
Correlational analyses showed that both intensity
and duration of ideation were significantly correlated
with variability and attempt status. Furthermore,
intensity was significantly correlated with duration.
Correlation coefficients for the three main parameters
of suicidality (i.e. intensity, duration, and variability)
can be found in Table 1.
1.7. Evaluating the unique relationship between the
parameters of suicidality and attempt status
Two separate, hierarchical regressions were conducted in order to determine which parameter, if any,
provided unique information for the prediction of
attempt status. Because intensity and duration were
correlated so highly with one another (r = .86), they
could not be entered simultaneously into a regression
due to collinearity problems. A dichotomous variable
indicating the presence of a mood disorder was
entered on Step 1 of all regressions because the
attempt status groups differed in regards to this variable. This variable was a significant predictor of previous attempt status in all regression equations. When
variability and duration were entered into the model
on Step 2, they accounted for 6% of the variance in
attempt status, although only variability was a significant predictor (see Table 2). A second regression
analysis was conducted with intensity and variability
entered Step 2. Similar results were found (see Table
2); only variability uniquely predicted attempt status.
Consistent with expectation, variability appears to be
the most potent predictor of attempt status. Moreover,

Table 1
Intercorrelations between parameters of suicidality (N = 108)
Parameter

SD

1.
2.
3.
4.

65.2
5.2
58.1

105.3
8.8
18.6

.24*

.30*
.86**

.28**
.22*
.23*

Variability
Duration
Intensity
Attempt status

Variability = Mean Squared Successive Difference of the Suicide

Probability Scale (SPS). Duration = number of days above a Tscore of 70 on the SPS. Intensity = mean score on the SPS.
* p b 05.
** p b .01.

T.K. Witte et al. / Journal of Affective Disorders 88 (2005) 131136

DV

IV

SE B

Attempt statusa

Mood disorder
Duration
Variability
Mood disorder
Intensity
Variability
Mood disorder
Gender
Variability
Gender  Variability

0.42
0.00
0.00
0.38
0.00
0.00
0.38
0.29
0.00
0.00

0.17
0.01
0.00
0.17
0.00
0.00
0.16
0.17
0.00
0.00

.24*
.07
.23*
.22*
.07
.23*
.22*
.19
.83**
.66*

Attempt statusb

Attempt statusc

DV = dependent variable. IV = independent variable.

a
R 2 = .09 for Step 1 ( p b .01); DR 2 = .06 for Step 2 ( p b .05).
b
R 2 = .08 for Step 1 ( p b .01); DR 2 = .06 for Step 2 ( p b .05).
c
R 2 = .08 for Step 1 ( p b .01); DR 2 = .06 for Step 2 ( p b .05);
DR 2 = .04 for Step 3 ( p b .05).
* p b .05.
** p b .01.

the other ideation parameters contributed no incremental or unique variance in explaining attempts.
1.8. Gender and the parameters of suicidality
Analysis of Variance (ANOVA) indicated no significant gender differences for variability ( F[1, 108] =
2.68, p = .11), duration ( F[1, 107] = 0.39, p = .53), or
intensity ( F[1, 108] = 0.02, p = .89). Three additional
regression analyses were conducted in order to determine if there was an interaction between gender and
any of the parameters of suicidality. The presence of a
mood disorder was entered on Step 1 of all of these
regressions as well. As Table 2 shows, variability was
a significant predictor of attempt status and accounted
for 6% of the total variance. As hypothesized, this
effect was qualified by a variability  gender interaction that explained an additional 4% of the variance.
We used Cohen and Cohens (1983, pp. 315317)
procedure to graph the interaction between gender
and variability in suicidal ideation, the results of
which are displayed in Fig. 1. Interestingly, variability
in suicidal ideation appeared to be less influential in
females than males. Males with a high MSSD scores
showed a propensity toward greater suicide attempts
compared to all other groups.
Neither the interaction of gender and duration (B =
0.02; SEB= 0.02;b = .12; p N .50) nor the interaction

of gender and intensity (B =  0.01; SE B = 0.01; b =

.53; p N .10) was a significant predictor of attempt
status.
2. Discussion
We demonstrated that variability is an important
parameter of suicidal ideation, and it provides better predictive power in terms of previous attempt
status relative to that provided by intensity or
duration. Perhaps this is because variability contains
aspects of both intensity and duration in addition to
some other component, which is likely to be mood
lability.
Crucially, however, the relation between attempt
status and variability in suicidal ideation appears to be
more important for male suicide attempters. Although
results did not demonstrate gender differences in
variability in suicidal ideation, the results were marginally significant ( p b .11), indicating that females
may have a tendency for higher levels of lability.
This is consistent with Winkler et al.s (2004) findings
that depressed females exhibit higher levels of mood
lability than do depressed males. It may be that
females, relative to males, are more likely to become
familiar with mood fluctuations or they may have
more opportunity to develop skills to cope with fluctuating moods. In males, the relative novelty of mood
fluctuation may create greater risk for an attempt,
given Zlotnick et al.s (1997) findings that affectively
unstable individuals use suicidal behavior to control
2

1.5
Attempt Status

Table 2
Summary of final steps in regression analyses the presence of a
mood disorder and parameters of suicidal ideation predicting
attempt status (N = 108)

135

0.5

Low MSSD

High MSSD

Variability in Suicidal Ideation

Males

Females

Fig. 1. Interaction of gender and variability in the prediction of

attempt status.

136

T.K. Witte et al. / Journal of Affective Disorders 88 (2005) 131136

moods. Emotionally labile males may also be more

likely to be exposed to painful and provocative stimuli
through risk-taking behaviors, putting them at greater
future risk for suicidal behavior.
There are several limitations of the current project
that should be noted. First, the sample of multiple
attempters was relatively small (n = 19), and of these,
only eight were male. Attempts to replicate these
results are certainly warranted. On the other hand, the
robust nature of the findings and their adherence to a
priori hypotheses suggest that these results are not
likely to be spurious. Another limitation is that the
sample consisted of young, college students. The suicide rate among college students is about half that of the
general population (Schwartz and Whitaker, 1990), so
it is possible that there are different mechanisms underlying suicidal behavior across these different samples.
On the other hand, the use of a young adult sample in
the current study is highly relevant to youth suicide,
which is important in its own right due to the increasing
incidence of youth suicide and the lack of research
relating to the parameters of youth suicide. In addition,
approximately half of the sample met current DSM-IV
criteria for a mood disorder, suggesting there is likely to
be some comparability with clinical samples. Finally,
the procedure utilized in the current study allows for the
possibility that participants may not have completed
the forms each day. However, we have no reason to
believe that those with different previous attempt status
would be more or less likely to deviate from instruction.
The current study has provided important information regarding the parameters of suicidal ideation. It
has also provided fertile ground for future explorations. Additional work should evaluate variability
prospectively and in other types of samples to fully
explore its predictive potential.

References
American Psychiatric Association, 1994. Diagnostic and Statistical
Manual of Mental Disorders, 4th ed. American Psychiatric
Association, Washington, DC.
Andrews, J.A., Lewinsohn, P.M., 1992. Suicidal attempts among
older adolescents: prevalence and co-occurrence with psychiatric
disorders. J. Am. Acad. Child Adolesc. Psych. 31, 655 662.
Beck, A.T., Steer, R.A., 1993. Beck Scale for Suicide Ideation. The
Psychological Corporation, San Antonio.

Beck, A.T., Brown, G.K., Steer, R.A., Dahlsgaard, K.K., Grisham,

J.R., 1999. Suicide ideation at its worst point: a predictor of
eventual suicide in psychiatric outpatients. Suicide Life-Threat.
Behav. 29, 1 9.
Cohen, J., Cohen, P., 1983. Applied Multiple Regression/Correlation Analysis for the Behavioral and Life Sciences, 2nd ed.
Erlbaum, Hillsdale, NG.
Cull, J.G., Gill, W.S., 1995. Suicide Probability Scale Manual.
Western Psychological Services, Los Angeles.
First, M.B., Spitzer, R.L., Gibbon, M., Williams, J.B., 1994. Structured Clinical Interview for Axis I DSM-IV DisordersPatient
Edition (SCID-I/P, Version 2.0). Biometrics Research Department, New York State Psychiatric Institute, New York.
Harvey, Greenberg, Serper, 1989. The affective lability scales: development, reliability, and validity. J. Clin. Psychol. 45, 786 793.
Joiner Jr., T.E., in press. Why People Die by Suicide. Cambridge,
MA: Harvard University Press.
Joiner, T., Rudd, M.D., 2000. Intensity and duration of suicidal
crises vary as a function of previous suicide attempts and
negative life events. J. Consult. Clin. Psychol. 68, 909 916.
Joiner Jr., T.E., Rudd, M.D., Rouleau, M.R., Wagner, K.D., 2000.
Parameters of suicidal crises vary as a function of previous
suicide attempts in youth inpatients. J. Am. Acad. Child Adolesc. Psych. 39, 876 880.
Lewinsohn, P.M., Rohde, P., Seely, J.R., 1996. Adolescent suicidal
ideation and attempts: prevalence, risk factors and clinical
implications. Clin. Psychol. Sci. Prac. 3, 25 46.
Neuringer, C., 1982. Affect configurations and changes in women
who threaten suicide following a crisis. J. Consult. Clin. Psychol. 50, 182 186.
Orbach, I., Stein, D., Shani-Sela, M., Har-Even, D., 2001. Body
attitudes and body experiences in suicidal adolescents. Suicide
Life-Threat. Behav. 31, 237 249.
Schwartz, A.J., Whitaker, L.C., 1990. Suicide among college students: assessment, treatment, and intervention. In: Blumenthal,
S.J., Kupfer, D.J. (Eds.), Suicide Over the Life Cycle: Risk
Factors, Assessment, and Treatment of Suicidal Patients. American Psychiatric Press, Washington, DC.
Townsend, E., Hawton, K., Harriss, L., Bale, E., Bond, A., 2001.
Substances used in deliberate self-poisoning 19851997: trends
and associations with age, gender, repetition and suicide intent.
Soc. Psychiatry Psychiatr. Epidemiol. 36, 228 234.
Winkler, D., Pjrek, E., Heiden, A., Wiesegger, G., Klein, N.,
Konstantinidis, A., Kasper, S., 2004. Gender differences in the
psychopathology of depressed inpatients. Eur. Arch. Psychiatry
Clin. Neurosci. 254, 209 214.
Witte, T.K., Fitzpatrick, K.K., Warren, K.L, Schatschneider, C.,
Schmidt, N.B., submitted for publication. Naturalistic evaluation
of suicidal ideation: variability and relation to attempt status.
Woyshville, M.J., Lackamp, J.M., Eisengart, J.A., Gilliland, J.A.M.,
1999. On the meaning and measurement of affective instability:
clues from chaos theory. Soc. Biol. Psychiatry 45, 261 269.
Zlotnick, C., Donaldson, D., Spirito, A., Pearlstein, T., 1997. Affect
regulation and suicide attempts in adolescent inpatients. J. Am.
Acad. Child Adolesc. Psych. 36, 793 797.