Conclusions
The overarching aim of this study was to harness participant-driven
descriptions of recent interpersonal stressors to better understand the
experiences of interpersonal difficulties as it relates to NSSI.
Specifically, we hypothesized that a range of categories of
interpersonal stress experiences would emerge. We also hypothesized that
NSSI history and emotion dysregulation would be associated with latent
topics in participant narratives of their interpersonal stressor (i.e.,
Victor et al., 2019; Turner, Wakefield et al., 2016). Finally, we
explored whether the emotional tone (i.e., valence) of the narrative was
related to the latent topics, or with NSSI and emotion dysregulation.
Regarding our first study aim, topic modeling indicated that a
four-topic model was optimal. As anticipated, a range of topics emerged,
including the interpersonal target, as reflected in Topic 2 (family
members) and 4 (friends, romantic partners). We also identified a topic
likely indicative of stressor content (Topic 1), which centered around
the discussion of daily or life difficulties (i.e., taxes, staffing,
homelessness). While a topic related to cognitive-affective or
behavioral responses to the interpersonal conflict was not apparent, a
topic related to speech-patterns, specifically involving the use of
adjectives and verbal-fillers (Topic 3), did emerge. These findings lend
some insight into both the targets and content of stressful
interpersonal interactions. Indeed, in highlighting a distinction
between family members and friends or romantic partners as targets in
the interpersonal interaction, findings are in line with prior research
demonstrating differences in daily communication patterns within
specific relationship types among those with a history of NSSI (Turner
et al., 2016)
The primary pattern of findings demonstrates greater use of language
consistent with Topic 3, which is most represented by adjective and
verbal-fillers, for undergraduate students with a history of NSSI and
higher levels of emotion dysregulation. Greater use of Topic 3 among
those with a history of NSSI and higher levels of emotion dysregulation,
regardless of sample, was also found. While this result was surprising,
it is possible that greater use of these “filler” words may be
representative of specific aspects of emotion regulation difficulties.
For example, the measure utilized to assess emotion dysregulation in
both studies is comprised of six dimensions, including lack of emotional
clarity and lack of emotional awareness. If any individual has
difficulty identifying their emotional reactions or arousal,
particularly when talking about a negatively-valanced event, they may be
more likely to use verbal fillers while trying to articulate their
experience. Work related to linguistics and NSSI is limited, but
findings within the suicide literature may lend insight. While research
related to verbal exchanges, specifically the predictive power of
acoustic properties of speech (i.e., France et al., 2000; for review see
Cummins et al., 2015) and properties of dyadic interactions (Nasir et
al., 2017) have been implicated in suicide risk, findings examining
online content from individuals at heightened suicide risk may have
greater relevance to the current findings. Indeed, it has been found
that online posts associated with greater suicide risk used more
quantifies, prepositions, and adverbs (Ji et al., 2018; O’Dea et al.,
2017). The current study is the first of our knowledge to extend these
results in relation to speech and among those with a history of NSSI.
Notably, we did not see the same pattern of finding related to Topic 3
and NSSI history and emotion dysregulation among the community sample.
Initially, we consider that this distinction may be driven by
differences in emotion dysregulation between the samples. However, the
community sample reported higher levels of emotion dysregulation (likely
due to overrecruitment for BPD features; Glenn & Klonsky, 2009) and the
three-way interaction between NSSI, emotion dysregulation, and sample
was not significant. Another potential explanation is differences in
valence of responses between samples; but we found that only those with
and without a history of NSSI differed on narrative valence. It is
possible that other sample characteristics (i.e., age, gender, education
level) may help explain our findings; however, such explorations are
beyond the scope of the current study and we encourage future research
to consider these relationships.
Consistent with the extraction of topics that distinguished between
family and friends / romantic partners, results demonstrated that among
individuals with a history of NSSI, there was a greater use of Topic 4
– which was largely represented by words related to friends and
romantic partners – as emotion dysregulation scores increased; but this
was not the case with the topic centered around family. These findings
may indicate that participants with a history of NSSI experience unique
stress in interpersonal relationships with friends or romantic partners.
Findings are in line with recent experience sampling research that found
NSSI urges were more likely to occur in close proximity to interactions
with friends and romantic partners as compared to parents or other
family members (Hepp et al., 2021). Moreover, prior research has
demonstrated that emotion regulation difficulties may mediate the
association between negative romantic relationship dynamics and NSSI
among college students (Silva et al., 2017).
Finally, in examining the emotional tone of participant dialogue, this
study showed that participants with a history of NSSI utilized more
negatively valanced language (such as “hardly”). This supports prior
research in the suicide literature, finding that online posts from
individuals at elevated suicide risk have different linguistic
properties, including greater use of negative emotion words (i.e., O’Dea
et al., 2017), as well as findings that those with a NSSI history have a
greater emotional response to interpersonal stressors (Kim et al.,
2015). Beyond the implications for emotional arousal, it may also be
that negatively valanced language is a way to undermine one’s own
description of a stressful event, which may be indicative of
self-invalidation. These findings would be consistent with associations
between NSSI and self-invalidation seen in prior research (Flett et al.,
2012). Greater work is needed in this area to better understand the
nuances of language among those with a NSSI history as a way to
disentangle such potential mechanisms.
This study is not without its limitations. First and foremost, both the
undergraduate and community samples are relatively homogenous
(predominantly white women), limiting the generalizability of findings
and our ability to examine the pattern of results by demographic
characteristics. Although we categorize our samples as an undergraduate
sample and a community sample, these samples cannot easily be
differentiated this way; there is a large proportion of full-time
students in the community sample, which may make it more difficult to
fully disentangle sample differences. On the other hand, there are also
notable differences between our samples that may impact findings. First,
the community sample was both larger with regard to sample size and
document length (i.e., number of words provided by each participant)
than the undergraduate sample. Second, the community sample was
oversampled for BPD features as part of a larger study, which resulted
in a significant difference in emotion dysregulation between the two
samples, with the community sample, regardless of NSSI status, reported
elevated emotion dysregulation scores. Finally, while participants were
asked to recall a recent interpersonal event, it is possible this
experience occurred weeks beforehand, limiting participant’s ability to
accurately recall their cognitive-affective experiences, which may be
why such a representative topic did not emerge; it may be beneficial to
consider a more proximal investigation of interpersonal events,
including situations that occurred shortly before a participant engaged
in NSSI. It will be important for future research to replicate these
findings, as well as probe some nuances, in a more diverse sample, both
with regard to demographics and NSSI history severity.
Despite these limitations, this study offers new insight into the
experiences of interpersonal stressors among those with and without a
history of NSSI. By using participant-driven dialogue and topic
modeling, we identified four topics related to stressful interpersonal
interactions and their unique relationships with clinically relevant
features (i.e., NSSI, emotion dysregulation). Notably, these findings
highlight a key distinction between relationships with one’s family, as
opposed to friends or romantic partners, with the latter seeming to hold
greater significance among those with a history of NSSI. Consequently,
future research examining interpersonal stress that also aims to
differentiate relationship context may be particularly important. We
also found differences in language use among those with a history of
NSSI; while these findings are in need of greater exploration, they
offer a novel avenue for assessment and identification of risk for NSSI.