Problematic scrolling climbed from secure to fearful attachment in a 555-adult US survey

People who both want closeness and fear relying on others reported the most problematic social media use in a PLOS One study of 555 US adults, published on October 21, 2026. Securely attached participants reported the least. Psychologist Chris Bjornsen and Allen Harbaugh, both of Longwood University in Virginia, ran the survey, and they say it cannot show cause.

Four attachment profiles, one ladder of scrolling

The study sorted its participants into four groups, using their answers on how anxious and how avoidant they feel in close relationships. The groups correspond to the secure, avoidant, anxious and fearful attachment styles, the four types the paper says were historically used to describe attachment. Other recent work on the same theme includes a Shanghai survey of 603 men, and an attachment theory expert's October 2 argument that no study has yet tested chatbot relationships against the conditions of attachment.

The paper defines them in plain terms. Secure attachment is a high degree of trust in close relationships. Anxious attachment is a strong desire for closeness with intense worry about whether others will be there. Avoidant attachment is discomfort with closeness and a preference for emotional distance. Fearful attachment combines a strong internal wish for closeness with a high level of fear of the possible negative consequences of relying on others.

Then came the comparison the authors had set out to make. Secure participants reported the lowest problematic social media use, and the avoidant, anxious and fearful groups reported increasingly higher levels, in that order. The differences were statistically significant between every pair of groups.

The authors define problematic use as impulsive, compulsive use meant to lift one's mood, despite negative consequences.

The ranking is also a hypothesis the paper states in its introduction: the secure group should report the lowest problematic use and the three insecure groups higher use. The authors expected attachment anxiety to matter more than avoidance, and they report that use was significantly higher in the anxious and fearful groups, the two that carry high anxiety. They add that they have no evidence of the causal relations between the variables.

Attachment was measured on two dimensions, anxiety and avoidance, with six questions each on a scale from 1 (strongly disagree) to 7 (strongly agree). The reliability figures reported for both scales were .90.

What the fearful profile looked like

The fearful group scored very high on attachment anxiety, very high on avoidance and very high on vulnerable narcissism, a trait the paper describes as a heightened need for admiration combined with shyness, shame about needing others and hypersensitivity to imagined criticism. This group also reported a very low level of interpersonal competence, which the study measured with a 15-item questionnaire about handling social situations.

The anxious group looked different. It scored high on anxiety and moderately low on avoidance, with moderately high scores on grandiose narcissism, vulnerable narcissism and interpersonal competence. The authors suggest that the common finding of a link between anxiety and poor social skills may apply mainly to people who are also high on avoidance, which describes the fearful group and not the anxious one.

The louder kind, grandiose narcissism, was not significantly linked to problematic social media use. The quieter vulnerable kind was.

On the reading of the authors, the highest scores belong to people who seek comfort for relationship distress online yet may find that contact too thin to reassure them. That is an interpretation, offered in the discussion section, and the survey did not test it.

Where earlier studies disagreed

The paper reviews a body of earlier work. Several systematic reviews and a meta-analysis found that heavy or addictive social media use is more consistently tied to anxious attachment than to avoidant attachment. Studies on the fearful style were fewer and less consistent: some found a low-to-moderate link with problematic smartphone use, others found none.

The authors propose a reason for the muddle. Many earlier studies did not classify anyone as fearful, so some participants may have been filed as avoidant. Sorting people into groups, rather than correlating two scores, is meant to avoid that confusion.

Their own avoidant group sits in the middle of the ladder: more problematic use than the secure group, less than the anxious and fearful groups. The authors read this as a sign that people who expect rejection but still want attention may find online contact less threatening and more controllable, given the ease with which one can dismiss or detach from online interactions.

The profiles also differed by gender and age, which the authors explored without making it their focus. There were more women than men in the anxious group, and more men than women in the secure one. Participants in the secure and avoidant groups were older than those in the anxious and fearful groups.

How the survey was built, and what it cannot say

The 555 participants were recruited from social media accounts and from two crowdsourcing platforms, Amazon MTurk and Prolific, restricted to US citizens, during spring and summer 2023. Of the original cases, 156 (22%) were dropped for incomplete answers or failed attention checks. Ages ran from 18 to 79, with a mean of 39.8. The questionnaire had 76 items and took about 13 minutes.

The measure of problematic use was built by the authors themselves: 10 original items written largely on the criteria the DSM-5, the American psychiatric manual, uses for alcohol use disorder. The paper's own limitations section says the validation would have been stronger with a previously validated scale such as the Bergen Social Media Addiction Scale.

The study is cross-sectional, meaning everyone answered once. The authors state that they have no evidence of the causal relations or the developmental sequence between attachment, narcissism, interpersonal skill and social media use, and say that longitudinal studies could illuminate it.

The sample is American, and the authors note that the results may not carry over to cultures that are less focused on individuality. Latent profile analysis, the method used to form the groups, has yet to become common, so they call for other samples to test whether the same four profiles appear. The paper reports no specific funding and no competing interests. The groups themselves are fuzzy, too. The model's entropy, a measure of how cleanly people fall into one group or another, was .646, which the authors describe as above the minimum recommended level but which leaves room for people near the borders.

What the researcher and a therapist take from it

Bjornsen told CNN that the study associated different attachment styles with different levels of problematic use but did not prove that certain styles cause the problem. In the same account, he said that people who are fearful about relationships might turn to social media to avoid interacting with other people.

He also described what he sees in his classroom: “Most of my students nowadays clearly will say, and own up to the fact, that they don’t know how to talk to each other in real life.” His advice to anyone whose use is interfering with work or school is to consider dropping social media altogether, or to curb it dramatically and see how they feel.

Kimberly Solo, a therapist in Medfield, Massachusetts, who was not involved in the research, put the offline side in terms of learning. Nurturing relationships is “a skill we learn, and it’s modeled for us as children.” Her practical suggestions are small: notice what you do to push people away, such as waiting days to answer a text, and set goals like learning to start a conversation.

CNN, which published the account on October 21, 2026, also notes that attachment styles form early, and that adults who lacked people to trust growing up may need to build social skills later. That is advice from two professionals, not a finding of the study.

For readers in the United States, the sample is the closest match. For readers in the United Kingdom, the paper offers no British data, so its profiles are a hypothesis to hold loosely, not a measurement of anyone's feed.

It may be worth noticing what the survey cannot hold: the slow part. Attachment, in this paper, is a score taken in a single sitting, while the bond it describes is built over many years of answered messages and shared meals. The authors name longitudinal studies as the way to learn the order.

The questionnaires were collected in spring and summer 2023, well before the paper appeared in October 2026.

Sources

• Attachment, grandiose and vulnerable narcissism, interpersonal competence, and problematic social media use: A latent profile analysis, PLOS One, Bjornsen and Harbaugh, October 21, 2026