The sound of loneliness: Prediction of perceived social isolation using automatic speech analysis

The sound of loneliness: Prediction of perceived social isolation using automatic speech analysis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article The sound of loneliness: Prediction of perceived social isolation using automatic speech analysis Diana Immel, Elisa Mallick, Nicklas Linz, Simon Barton, Rene Hurlemann, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7868485/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 04 Apr, 2026 Read the published version in Scientific Reports → Version 1 posted 8 You are reading this latest preprint version Abstract Loneliness has been demonstrated to exert a detrimental effect on mental and physical health. It may impede the formation of new social relationships by altering interactional behavior. This study provides a proof of concept that loneliness is reflected in altered speech, demonstrating that small yet significant effects can make loneliness audible. Samples of 96 healthy participants (mean age 30.85 years, 53 women) were recorded while they performed a picture description and storytelling task. Paralinguistic markers related to prosodic, formant, source, and temporal qualities of speech were extracted and correlated with loneliness, social anxiety and depression. To validate the diagnostic power, machine learning analyses were conducted for women and men separately. A model comprising all speech features from the picture description task significantly predicted loneliness. However, this model did not predict loneliness from the storytelling task. No single speech feature emerged as a strong predictor of loneliness. A combined model that included both speech features and psychiatric symptoms provided better predictions than psychiatric symptoms alone only in women. Overall, these findings suggest that speech offers a new perspective on how loneliness becomes perceptible to others and how it may disrupt social interactions, thereby fostering chronicity. Health sciences/Health care Biological sciences/Psychology Social science/Psychology depression loneliness speech biomarker social anxiety speech analysis Figures Figure 1 Introduction Human beings are social by nature and are intrinsically motivated to form and maintain interpersonal relationships. When a person's need to belong is not consistently met, feelings of loneliness develop. Loneliness has a cross-age prevalence rate of up to 33% and has detrimental effects on mental and physical health. It is associated with an increased risk of psychiatric disorders such as depression [ 1 ], psychosis [ 2 ], and social anxiety [ 3 ]. Importantly, it is also associated with suicidality [ 4 ] and higher all-cause mortality [ 5 ]. Multiple lines of evidence indicate that loneliness is associated with negative cognitive biases [ 6 , 7 , 8 ]. For instance, implicit hypervigilance to social threats, increased anticipation of rejection, and negative attributional styles can result in higher expectations of negative social interactions. At the neural level, reduced reactivity and functional connectivity of the anterior insula may reflect impaired integration of trust-related information in lonely individuals [ 9 ]. Despite the phenotypic overlap, the neural mechanisms associated with loneliness appear to be distinct from those associated with social anxiety [10]. Importantly, lonely individuals tend to elicit behaviors from interaction partners that lead to the confirmation of their negative expectations, resulting in a self-fulfilling prophecy [ 6 , 11 ]. This theoretical framework also helps to explain why subtle alterations in speech may emerge as markers of loneliness because communicative behavior reflects and reinforces social expectations. Consequently, lonely individuals prefer greater social distance in interactions with strangers and benefit less from positive social interactions [9, 12]. Meta-analytical evidence also indicates a small negative association between loneliness and prosociality [ 7 ]. Although several studies support the notion that lonely individuals have impaired social interactions, the underlying behavioral mechanisms remain elusive. Preliminary evidence suggests that loneliness is associated with emotion-specific impairments in the recognition of vocal expressions [ 13 ]. However, given the bidirectional nature of social interactions, it is conceivable that both altered speech recognition and production may result in dysfunctional social interactions. Previously, we found that the hypothalamic peptide oxytocin facilitates communicative reciprocity by enhancing the salience of vocal expressions [ 14 ]. Additionally, we found that interaction-induced oxytocin release is significantly impaired in lonely individuals [ 9 ]. These oxytocin-related deficits provide a theoretical basis for why loneliness may be detectable through paralinguistic speech features. Consistent with this, self-reports of communicative competence negatively correlate with loneliness [15]. Notably, machine learning experiments are increasingly being used to identify speech patterns in patients with mental disorders. Automated speech analysis has been employed to detect apathy in older adults with cognitive impairments and to classify patients with affective disorders [ 16 ], posttraumatic stress disorder [ 17 ], and suicidal ideation [ 18 ]. A recent study found that machine learning models based on natural language processing of manually transcribed qualitative interviews can be used to predict loneliness in older adults [ 19 ]. However, given the emotion-specific effects of loneliness on speech production, it is still unclear whether changes in speech features associated with loneliness are more pronounced in emotional content and in older individuals, and whether a dyadic setting is required. To this end, we used the revised UCLA Loneliness Questionnaire (UCLA) [20, 21] to assess loneliness in 96 healthy participants (53 women, mean age 30.85 years; 43 men, mean age 31.37 years). We recorded the participants’ speech during three tasks. They described the Cookie Theft picture from the Boston Diagnostic Aphasia Examination [ 22 ] and were asked to talk about positive and negative life events [ 23 ]. We used Random Forest regression models to predict the UCLA score from extracted acoustic speech features. Because speech features naturally vary by sex, we conducted the analyses separately for women and men. Due to the phenotypic overlap between loneliness, depression, and social anxiety, we also assessed depression and social anxiety as control variables. We hypothesized that loneliness would be associated with significantly altered paralinguistic markers relating to the prosodic, formant, source, and temporal qualities of speech [ 24 ]. Results We observed average loneliness scores of 41.72 in women and 38.02 in men (see Table 1 ) which are comparable to the normative values reported for the UCLA Loneliness Scale in college students [ 20 ]. There were no significant differences in loneliness, depression, or social anxiety between women and men (see Table 1 ). In both genders, loneliness significantly correlated with depression (women: r (53) = 0.59, p < .001; men: r (43) = 0.54, p < .001) and social anxiety (women: r (53) = 0.34, p < .005; men: r (43) = 0.32, p 0.9; see Table 1 ). Table 1 Sample description Females M ( SD ) Males M ( SD ) U (96) p Cronbachs Alpha Age (years) 30.85 (10.90) 31.37 (9.97) 1246.10 .43 - Education (years) 16.28 (2.82) 16.53 (2.42) 1182.50 .75 - Loneliness 41.72 (14.71) 38.02 (13.34) 934 .13 .96 Depression 6.13 (6.29) 6.98 (7.88) 1186.5 .73 .90 Social Anxiety 15.85 (10.86) 12.14 (11.15) 862.50 .04 .95 Importantly, the set of speech features extracted from the picture description task by machine learning models significantly predicted loneliness scores of both women and men (see Fig. 1 ). The model explained 6% and 16% of the variance in women and men, respectively (see Table 2 ). As expected, depression and social anxiety also predicted loneliness in the machine learning analyses. For women, a combined model that included speech features from the picture description, depression, and social anxiety scores as predictors explained more variance ( R 2 = .21, p < .01) than reduced models that included either speech features from the picture description ( R 2 = .06, p = .04) or depression and social anxiety scores ( R 2 = .16, p = .02). However, in men, the combined model with speech features, depression, and social anxiety scores ( R 2 = .16, p = .03) resulted in a worse fit than the reduced model without the picture description speech features ( R 2 = .28, p < .01) (see Table 2 ). Table 2 Results of machine learning experiments for loneliness (picture description) Females Males Features set Random Forest Randomised Baseline Random Forest Randomised Baseline MAE R 2 M ( SD ) p MAE R 2 M ( SD ) p Picture Description 11.49 .06 13.39 (0.95) .04 10.25 .16 12.50 (0.92) .02 Depression + Social Anxiety 11.24 .16 14.35 (1.48) .02 8.96 .28 13.10 (1.47) < .01 Picture Description + Depression + Social Anxiety 10.47 .21 13.36 (0.95) < .01 10.10 .16 12.51 (0.91) .03 Baseline 13.25 12.08 Notes. MAE: Mean absolute error of random forest regression. Not surprisingly, no single speech feature emerged as a strong predictor of loneliness. For picture description, significant correlations between extracted speech features and loneliness were evident in the temporal and source categories (see Table 3 ). Specifically, higher loneliness was significantly associated with a lower speech to non-speech ratio in women. Similarly, a higher kurtosis value in the amplitude distribution of the signal from the picture description task was associated with greater loneliness. This suggests that volume intensity was distributed more irregularly among lonely women. In men, greater loneliness significantly correlated with fewer pauses between syllables and a shorter phonation time. Additionally, men with higher loneliness scores had a lower sound-to-noise ratio, reflecting poorer voice quality, and a higher mean pitch. Table 3 TOP 5 highest spearman rank partial correlations between speech features and loneliness, for the picture description Females Speech features Speech ratio Amplitude kurtosis Peak frequency Amplitude mean absolute value Mean power Coefficient − .29 .28 − .22 − .21 .13 p .03 .04 .12 .14 .34 Males Speech features Number of pauses Total phonation time Sound to noise ratio Harmonics to noise ratio Power spectrum ratio Coefficient − .44 − .43 − .37 − .25 − .21 p < .01 < .01 .02 .10 .17 The set of speech features extracted from the free speech emotional storytelling tasks did not significantly predict loneliness in either positive or negative storytelling for women or men. As expected, the combined models (negative/positive storytelling + depression + social anxiety) explained more variance than the reduced models that examined only storytelling for both genders. However, the combined model only became significant ( R 2 = .09, p = .04) for negative storytelling among women (see Tables S1, S2 ). Discussion In the present study, we examined whether loneliness is reflected in speech features in a heterogeneous sample of young healthy adults. Using a machine learning-based statistical approach, we found that paralinguistic markers extracted from a semi-guided picture description task significantly predicted loneliness in both women and men. Speech features from the temporal and source categories appear to be particularly relevant to this association. Interestingly, a model that included both speech features and depression and social anxiety scores enabled a better prediction than a model only with psychiatric symptoms in women, but not men. However, extraction of speech features from positive and negative free storytelling did not significantly predict loneliness. Loneliness can affect social interactions in numerous ways. For instance, highly lonely individuals prefer greater distance from an unfamiliar interaction partner [ 9 ], exhibit altered gaze processing [ 25 ], and increased gaze towards their conversation partners' faces [ 26 ]. Sleep-deprived participants have been rated as significantly lonelier and less desirable to interact with [27]. Furthermore, blinded experimenters were able to identify whether they were interacting with a lonely or non-lonely individual [ 9 ]. A previous study found that loneliness could be predicted from the content of transcribed speech using natural language processing in older adults [ 19 ]. However, our findings suggest that loneliness is also reflected in paralinguistic markers. This highlights an innovative shift from “what is said” to “how it is said.” Consistent with the multifaceted nature of loneliness, the present proof-of-concept study builds on previous approaches of natural language processing by demonstrating that alterations in speech related to loneliness are not limited to linguistic content, but can also be detected in paralinguistic domains (e.g., temporal, source-related, spectral or prosodic speech categories). Interestingly, speech features extracted from emotional storytelling did not significantly predict loneliness. Arousal-induced speech changes may have obscured loneliness-specific markers. Another possibility is that the Cookie Theft scenario triggered stronger social-cognitive processing because participants had to infer intentions, roles, and relationships between characters. These demands may directly activate biases related to loneliness in attention and interpretation, which could be reflected in paralinguistic speech markers. In contrast, free storytelling about personal life events may rely more on autobiographical memory and emotional arousal. These processes could overshadow subtle, loneliness-related alterations in speech. Previously, a reduced oxytocinergic response to semi-guided social interactions was observed in individuals with high loneliness [ 9 ]. Reduced oxytocin release may impair the transmission of emotional information in social settings because exogenous (e.g. nasally administered) oxytocin enhanced facial and vocal expression of fear and happiness [ 14 ]. These observations highlight that the detectability of loneliness-related speech markers is likely task- and context-dependent. Further research is also needed to investigate whether changes in speech features are related to endocrine function. Although loneliness is an important risk factor for depression and anxiety, accumulating evidence suggests that it should be considered a distinct construct. In a prospective longitudinal study, loneliness predicted subsequent changes in depressive symptomatology, but not vice versa [28]. Similarly, loneliness exhibits a unique neural profile during cognitive control tasks in patients with major depressive disorder and in healthy controls [ 29 ]. Additionally, evidence suggests that lonely and non-lonely individuals experience equal subjective valence when engaging in social situations, as well as exhibit comparable amygdala responses to social decisions and striatal responses to positive social feedback [10]. This pattern of responses stands in stark contrast to the findings for social anxiety [ 30 ]. In the present study, loneliness significantly correlated with depression and social anxiety in both women and men. Interestingly, the combined predictive model, which included speech features from the Picture Description Task, as well as depression and social anxiety scores, provided a better model fit for females. For males, however, this model showed a poorer fit than the reduced models, a tendency also evident in the storytelling tasks. These results suggest that loneliness may follow gender-specific pathways. Prior studies support such differences. Specifically, loneliness has been associated with a more pronounced within-network coupling of the default network in men than in women [ 31 ]. An interaction between loneliness and gender was also found following an experimental trauma paradigm: more intrusions were reported by lonely men, but not by lonely women [ 32 ]. This effect was accompanied by reduced amygdala habituation to repeated fearful faces and amygdala hyperreactivity during fear conditioning in lonely men. Our results contribute to the existing literature by suggesting that the prediction of loneliness through speech may be more strongly moderated by comorbid symptoms in men than in women. While the driving mechanisms remain unclear, emphasizing gender as a potential moderator is an important direction for future hypothesis-driven research on sex-specific pathways of social communication. There are several limitations to the current study. First, we recruited healthy individuals with varying levels of loneliness, so it is unclear whether our findings can be generalized to patients with depression or anxiety disorders. Additionally, chronic loneliness is a relatively stable construct with trait-like properties [33]. However, it is likely that the adverse health consequences of loneliness depend on its chronicity [ 34 ]. Even brief periods of social isolation can lead to decreased energy levels and increased feelings of fatigue [ 35 ], but situational loneliness seems to drive people toward reconnection, while chronic loneliness seems to drive people away from it [12]. We assessed trait-like loneliness using the established UCLA Loneliness Scale, which does not allow conclusions about the chronicity of the perceived social isolation. Taken together, these findings provide the first evidence that loneliness can be predicted by paralinguistic markers that are automatically extracted from semi-guided speech. This mechanism may explain why loneliness can be perceived by others and shed light on a pathway by which loneliness may hinder positive interactions, thereby propagating the maintenance of chronic loneliness. Future research should test these approaches in larger, more diverse samples, including clinical populations, and adopt longitudinal designs that capture loneliness dynamics over time. Furthermore, incorporating speech-based assessments into ecologically valid settings, such as everyday social interactions or digital health platforms, could substantially increase their translational potential. Methods 4.1 Participants Eligibility for the study included the following requirements: Participants had to be between 18 and 65 years of age, speak sufficient German, and not have a psychiatric diagnosis or be taking psychiatric medication. Two sources were used to recruit participants. First, healthy participants were recruited through online advertisements and public notices. Second, pre-stratified healthy participants in a group therapy intervention aimed at reducing loneliness were asked to participate in the study before starting the therapy intervention. A total of 105 participants were included in the study. Nine participants were excluded from the analyses due to missing voice recordings or other missing data. The final sample consisted of N = 96 people (53 women and 43 men). The mean age was 30.85 years (± SD : 10.90) for women and 31.37 years (± SD : 9.97) for men. The study was approved by the Ethics Committee of the University Hospital of Bonn and was conducted according to the principles of the Declaration of Helsinki. The study and data analyses were pre-registered ( https://osf.io/buqrj/ ). Participants were enrolled after providing written informed consent and received monetary compensation at the end of the study. 4.2 Study Tasks The “Cookie Theft” picture from the Boston Diagnostic Aphasia Examination is a well-established method for assessing the expressive language skills of children and adults [ 36 ]. One feature of the task is that it elicits mental state language [ 37 ]. The picture depicts a familiar domestic scene that requires making assumptions about the mental states of the characters. For instance, the mother is daydreaming and therefore does not notice her children climbing on a stool that is about to fall while they scramble for biscuits. In the free emotional storytelling task, the participants were asked to talk about a negative and a positive event in their lives [ 23 ]. 4.3 Questionnaires In addition to the speech assessment, clinical measures were also collected as part of the investigation. The Becks Depression Inventory (BDI-II) is a psychological self-report instrument (21 items with a 4-point Likert scale) for assessing the severity of depression in adolescents and adults ranging from 0 to 63 [ 38 ]. The Liebowitz Social Anxiety Scale (LSAS) (50 items with a 4-point Likert scale) is a questionnaire with a range from 0 to 72 used for the diagnosis of social anxiety disorder [ 39 ]. Trait-like loneliness was measured using a validated German version of the Revised UCLA Loneliness Scale (UCLA) [40], which is a 20-item, 4-point Likert scale with scores ranging from 20 to 80. Numerous validation studies have established loneliness as a distinct psychological construct [ 20 ], [41]. Psychometric test properties, such as retest reliability and internal consistency, are considered satisfactory [ 42 , 43 ]. 4.4 Procedure All participants attended one 125-minute study session. The objective of the study and the study procedure were explained. The inclusion and exclusion criteria were explicitly assessed, and written consent was requested before the assessment began. All participants were screened using the Mini-International Neuropsychiatric Interview (MINI) [ 44 ]. Then, psychometric questionnaires were administered using Qualtrics software (Provo, USA). Then, the speech assessment was administered on an Apple iPad tablet performed by the ∆elta Clinical app [ 45 , 46 ]. This study was part of a larger study, the results of which are described elsewhere. The speech assessment took approximately five minutes per task and was conducted with an experimenter present. During the speech tasks, the tablet recorded the participants' speech features. 4.5 Data Analysis The speech data consist of various speech features (see Table S8 , that were automatically extracted from the audio signal by the iOS app ∆elta Clinical [ 46 ]). These features were extracted separately for picture description, positive storytelling, and negative storytelling. They are grouped into four main categories: Temporal features indicate the general rate of speech and measure the proportion of speech (e.g., length and connection of speech segments and the pauses between them). These features reflect the effectiveness of speech production and overlap with prosodic speech characteristics, in the form of fluency and rhythm. Prosodic features refer to the long-term dynamics of perceived intonation and speech rhythm. These features demonstrate the overall speech melody adapted to a given situation, thereby indicating prosodic competence in terms of appropriate of speech intonation [47, 48]. Prosodic features also measure changes in an individual's speaking style (e.g., perceived intonation or pitch). Spectral features represent the relationship between articulatory movements and changes in vocal tract shape. These features include spectral flow, energy, slope and flatness [ 49 ]. Spectral features measure the airborne noise caused by the speech signal and the power of the strongest frequency relative to all others, such as background noise, which can be filtered out to improve speech analysis. Source features are important markers of voice quality. They represent the auditory perceptibility of changes in vocal fold vibration and vocal tract shape, outside of pitch, loudness, and phonetics. Source features frequently record information about laryngeal qualities, such as breathing, creaking, hardness, and phonation type [50]. Due to noise (e.g., background noise) in the processing, the speech features espinola zero crossing metric, mean F0, and average amplitude change demonstrated zero values and were excluded from the analysis. 4.5.1 Statistical Analysis The database contains 78 speech features and three clinical scores (UCLA, BDI-II and LSAS) from 105 individuals. To account for gender-specific differences in speech characteristics between women and men [51], the data were analyzed separately by gender. QQ plots and Shapiro-Wilk tests revealed that the UCLA, BDI-II and LSAS scores did not follow a normal distribution. The total LSAS score was obtained by summing the anxiety and avoidance subscores. During the data analysis process, it was decided to deviate from the originally planned registration. The variable "loneliness" was used as a continuous variable to mitigate the loss of information. Consequently, the area under the curve calculation, which is used to make predictions, was not conducted. Statistical analysis was performed with Rstudio (version 1.4.1103). Spearman rank correlations were calculated between UCLA, BDI-II and LSAS scores and speech characteristics for both sexes. Demographic and psychological variables were compared between sexes with Mann-Whitney-U-tests. Internal consistency of the three clinical scores (UCLA, BDI-II, and LSAS) was calculated (IBM SPSS Statistics (Version 30)) using Cronbach's alpha, a statistical method for measuring internal consistency in scales or inventories. 4.5.2 Machine learning experiments Random forest regression models were used to predict the UCLA score based on acoustic speech features extracted from speech tasks, as well as the BDI-II and LSAS scores. The features were normalized using a standard scaler. The models were trained using leave-one-out cross-validation and grid search for hyperparameter tuning. Mean absolute error (MAE) and R-squared are reported as performance measures. The model results were compared with the baseline MAE obtained by predicting the population mean. To calculate the statistical significance of the regression models' performance, a randomized baseline was used, consisting of training an extra tree model several times with the labels permuted each time. 4.5.3 Power Analyses To date, no study has examined the relationship between loneliness and automatically extracted speech features. Therefore, an a priori power analysis was conducted for this project using G*Power 3. This analysis was based on the effect size obtained in a previous study that examined the effects of loneliness on affective responsiveness to a positive social interactions [ 9 ]. The results showed that the positive mood change induced by an interaction was significantly reduced in participants with high loneliness ( r (79) = − .25, p = .03). To reliably replicate this effect of loneliness (with α = .05, and power = .80, one-tailed t -test), at least 95 participants must be tested. To account for possible dropouts, the plan was to test at least 100 participants (50 women). Declarations Conflict of interest statement Elisa Mallick is employed by the company ki:elements, which developed the application for the speech-based assessment and extracted the speech features. Nicklas Linz owns shares in the ki:elements company. Dirk Scheele, Simon Barton, Rene Hurlemann and Diana Immel have nothing to disclose. Funding sources R.H. and D.S. were supported by a grant from the German Research Foundation (DFG) (HU 1302/18-1 and SCHE 1913/7-1). Authors' contributions D.I. and D.S. designed the experiment; D.I. performed the experiments; D.I., E.M., S.B. and N.L. analysed the data. All authors drafted the manuscript. All authors read and approved the current version of the manuscript. Data availability statement The data will be provided upon reasonable request. References Cacioppo, J. T., Hughes, M. E., Waite, L. J., Hawkley, L. C. & Thisted, R. A. Loneliness as a specific risk factor for depressive symptoms: Cross-sectional and longitudinal analyses. Psychol. Aging 21 , 140–151 (2006). doi: 10.1037/0882-7974.21.1.140 Velthorst, E. et al. The impact of loneliness and social relationship dissatisfaction on clinical and functional outcomes in Dutch mental health service users. Psychiatry Res. 342 , 116242 (2024). doi: 10.1016/j.psychres.2024.116242 Maes, M. et al. Loneliness and social anxiety across childhood and adolescence: Multilevel meta-analyses of cross-sectional and longitudinal associations. Dev. Psychol. 55 , 1548–1565 (2019). doi: 10.1037/dev0000745 Institute of Medicine. Reducing Suicide: A National Imperative . (National Academies Press, 2002). doi: 10.17226/10398 Holt-Lunstad, J., Smith, T. B., Baker, M., Harris, T. & Stephenson, D. Loneliness and social isolation as risk factors for mortality: a meta-analytic review. Perspect. Psychol. Sci. 10 , 227–237 (2015). doi: 10.1177/1745691614568352 Hawkley, L. C. & Cacioppo, J. T. Loneliness matters: A theoretical and empirical review of consequences and mechanisms. Ann. Behav. Med. 40 , 218–227 (2010). doi: 10.1007/s12160-010-9210-8 Malon, M., Gajos, K., Rajchert, J., Holt-Lunstad, J. & Okruszek, Ł. Lonely and self-centered? A meta-analysis of the link between prosociality and loneliness. Pers. Soc. Psychol. Bull. (2024). doi: 10.1177/01461672241295263 Spithoven, A. W. M., Bijttebier, P. & Goossens, L. It is all in their mind: A review on information processing bias in lonely individuals. Clin. Psychol. Rev. 58 , 97–114 (2017). doi: 10.1016/j.cpr.2017.10.003 Lieberz, J. et al. Loneliness and the social brain: How perceived social isolation impairs human interactions. bioRxiv (2021). doi: 10.1002/advs.202102076 Cacioppo, J. T., Norris, C. J., Decety, J., Monteleone, G. & Nusbaum, H. In the eye of the beholder: Individual differences in perceived social isolation predict regional brain activation to social stimuli. J. Cogn. Neurosci. 21 , 83–92 (2009). doi: 10.1162/jocn.2008.21007 Saporta, N. et al. Opposing association of situational and chronic loneliness with interpersonal distance. Brain Sci. 11 , 1135 (2021). doi: 10.3390/brainsci11081135 Morningstar, M., Nowland, R., Dirks, M. A. & Qualter, P. Loneliness and the recognition of vocal socioemotional expressions in adolescence. Cognition Emot. 34 , 970–976 (2020). doi: 10.1080/02699931.2019.1688602 Spengler, F. B. et al. Oxytocin facilitates reciprocity in social communication. Soc. Cogn. Affect. Neurosci. 12 , 1325–1333 (2017). doi: 10.1093/scan/nsx056 Zakahi, W. R. & Duran, R. L. Loneliness, communicative competence, and communication apprehension: Extension and replication. Commun. Q. 33 , 50–60 (1985). doi: 10.1080/01463378509369696 Flanagan, O., Chan, A., Roop, P. & Sundram, F. Using acoustic speech patterns from smartphones to investigate mood disorders: Scoping review. JMIR Mhealth Uhealth 9 , e24352 (2021). doi: 10.2196/24352 Marmar, C. R. et al. Speech-based markers for posttraumatic stress disorder in US veterans. Depress. Anxiety 36 , 607–616 (2019). doi: 10.1002/da.22946 Belouali, A. et al. Acoustic and language analysis of speech for suicidal ideation among US veterans. BioData Mining 14 , 11 (2021). doi: 10.1186/s13040-021-00245-0 Badal, V. D. et al. Prediction of loneliness in older adults using natural language processing: Exploring sex differences in speech. Am. J. Geriatr. Psychiatry 29 , 853–866 (2021). doi: 10.1016/j.jagp.2021.02.017 Russell, D., Peplau, L. A. & Cutrona, C. E. The revised UCLA Loneliness Scale: Concurrent and discriminant validity evidence. J. Pers. Soc. Psychol. 39 , 472–480 (1980). doi: 10.1037/0022-3514.39.3.472 Schwab, R. Zur Messung und klinisch-psychologischen Bedeutung von Einsamkeit: Eine deutschsprachige Version der „Rev. UCLA Loneliness Scale“ von Russell, Peplau & Cutrona (1980). In Bericht über den 13. Kongreß für Angewandte Psychologie , Bd 2, 245–267 (Bonn, 1985). Giles, E., Patterson, K. & Hodges, J. R. Performance on the Boston Cookie Theft picture description task in patients with early dementia of the Alzheimer’s type: Missing information. Aphasiology 10 , 395–408 (1996). doi: 10.1080/02687039608248460 König, A. et al. Detecting apathy in older adults with cognitive disorders using automatic speech analysis. J. Alzheimers Dis. 69 , 1183–1193 (2019). doi: 10.3233/JAD-180947 Bell, V. et al. Do loneliness and social exclusion breed paranoia? An experience sampling investigation across the psychosis continuum. Schizophr. Res. Cogn. 33 , 100282 (2023). doi: 10.1016/j.scog.2023.100282 Kanai, R. et al. Brain structure links loneliness to social perception. Curr. Biol. 22 , 1975–1979 (2012). doi: 10.1016/j.cub.2012.08.030 Lodder, G. M. A., Scholte, R. H. J., Goossens, L., Engels, R. C. M. E. & Verhagen, M. Loneliness and the social monitoring system: Emotion recognition and eye gaze in a real-life conversation. Br. J. Psychol. 107 , 135–153 (2016). doi: 10.1111/bjop.12114 Ben Simon, E. & Walker, M. P. Sleep loss causes social withdrawal and loneliness. Nat. Commun. 9 , 3146 (2018). doi: 10.1038/s41467-018-05634-1 Cacioppo, J. T., Hawkley, L. C. & Thisted, R. A. Perceived social isolation makes me sad: 5-year cross-lagged analyses of loneliness and depressive symptomatology in the Chicago Health, Aging, and Social Relations Study. Psychol. Aging 25 , 453–463 (2010). doi: 10.1037/a0017216 Shao, R. et al. Loneliness and depression dissociated on parietal-centered networks in cognitive and resting states. Psychol. Med. 50 , 2691–2701 (2020). doi: 10.1017/S0033291719003096 Schultz, J. et al. A human subcortical network underlying social avoidance revealed by risky economic choices. eLife 8 , e45249 (2019). doi: 10.7554/eLife.45249 Spreng, R. N. et al. The default network of the human brain is associated with perceived social isolation. Nat. Commun. 11 , 6393 (2020). doi: 10.1038/s41467-020-20243-0 Morr, M. et al. Lonely in the dark: Trauma memory and sex-specific dysregulation of amygdala reactivity to fear signals. Adv. Sci. 9 , 2105336 (2022). doi: 10.1002/advs.202105336 Mund, M., Freuding, M. M., Möbius, K., Horn, N. & Neyer, F. J. The stability and change of loneliness across the life span: A meta-analysis of longitudinal studies. Pers. Soc. Psychol. Rev. 24 , 24–52 (2020). doi: 10.1177/1088868320902698 Haihambo, N., Layiwola, D.-M., Blank, H., Hurlemann, R. & Scheele, D. Loneliness and social conformity: A predictive processing perspective. Ann. N. Y. Acad. Sci. 1547 , 5–17 (2025). doi: 10.1111/nyas.14853 Stijovic, A. et al. Homeostatic regulation of energetic arousal during acute social isolation: Evidence from the lab and the field. Psychol. Sci. 34 , 537–551 (2023). doi: 10.1177/09567976221138099 Goodglass, H., Kaplan, E. & Weintraub, S. BDAE: The Boston Diagnostic Aphasia Examination . Lippincott Williams & Wilkins, Philadelphia, PA (2001). Cummings, L. Describing the Cookie Theft picture: Sources of breakdown in Alzheimer’s dementia. Pragmatics Soc. 10 , 153–176 (2019). doi: 10.1075/ps.18023.cum Beck, A. T., Steer, R. A. & Brown, G. K. Beck Depression Inventory (BDI-II) . vol. 10, Pearson, London, UK (1996). Rytwinski, N. K. et al. Screening for social anxiety disorder with the self-report version of the Liebowitz Social Anxiety Scale. Depress. Anxiety 26 , 34–38 (2009). doi: 10.1002/da.20508 Russell, D., Peplau, L. A. & Ferguson, M. L. Developing a measure of loneliness. J. Pers. Assess. 42 , 290–294 (1978). doi: 10.1207/s15327752jpa4203_11 Weeks, D. G., Michela, J. L., Peplau, L. A. & Bragg, M. E. Relation between loneliness and depression: A structural equation analysis. J. Pers. Soc. Psychol. 39 , 1238–1244 (1980). doi: 10.1037/0022-3514.39.6.1238 Hartshorne, T. S. Psychometric properties and confirmatory factor analysis of the UCLA Loneliness Scale. J. Pers. Assess. 61 , 182–195 (1993). doi: 10.1207/s15327752jpa6101_14 Russell, D. W. UCLA Loneliness Scale (Version 3): Reliability, validity, and factor structure. J. Pers. Assess. 66 , 20–40 (1996). doi: 10.1207/s15327752jpa6601_2 Sheehan, D. V. et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): Development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J. Clin. Psychiatry 59 Suppl 20 , 22–33; quiz 34–57 (1998). Baykara, E., Kuhn, C., Linz, N., Tröger, J. & Karbach, J. Validation of a digital, tablet-based version of the Trail Making Test in the ∆elta platform. Eur. J. Neurosci. 55 , 461–467 (2022). doi: 10.1111/ejn.15191 Linz, N. et al. Automatic detection of apathy using acoustic markers extracted from free emotional speech. In Proc. 40th Annu. Conf. Cogn. Sci. Soc. , 1849–1855 (Cognitive Science Society, Austin, TX, 2018). Aloshban, N., Esposito, A. & Vinciarelli, A. What you say or how you say it? Depression detection through joint modeling of linguistic and acoustic aspects of speech. Cogn. Comput. 14 , 1585–1598 (2022). doi: 10.1007/s12559-022-09934-4 Wade-Woolley, L., Wood, C., Chan, J. & Weidman, S. Prosodic competence as the missing component of reading processes across languages: Theory, evidence and future research. Sci. Stud. Read. 26 , 165–181 (2022). doi: 10.1080/10888438.2022.2028244 Wu, P. et al. Automatic depression recognition by intelligent speech signal processing: A systematic survey. CAAI Trans. Intell. Technol. 8 , 701–711 (2023). doi: 10.1049/trit.2022.0053 Gobl, C. & Ní Chasaide, A. The role of voice quality in communicating emotion, mood and attitude. Speech Commun. 40 , 189–212 (2003). doi: 10.1016/S0167-6393(02)00059-3 Cummins, N., Vlasenko, B., Sagha, H. & Schuller, B. Enhancing speech-based depression detection through gender dependent vowel-level formant features. In Artificial Intelligence in Medicine , Lecture Notes in Computer Science, vol. 12501, 209–214 (Springer, Cham, 2020). doi: 10.1007/978-3-030-58814-5_26 Additional Declarations Competing interest reported. Elisa Mallick is employed by the company ki:elements, which developed the application for the speech-based assessment and extracted the speech features. Nicklas Linz owns shares in the ki:elements company. Dirk Scheele, Simon Barton, Rene Hurlemann and Diana Immel have nothing to disclose. Supplementary Files SISoL.docx Cite Share Download PDF Status: Published Journal Publication published 04 Apr, 2026 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 31 Dec, 2025 Reviews received at journal 15 Dec, 2025 Reviewers agreed at journal 17 Nov, 2025 Reviewers invited by journal 30 Oct, 2025 Editor invited by journal 17 Oct, 2025 Editor assigned by journal 16 Oct, 2025 Submission checks completed at journal 16 Oct, 2025 First submitted to journal 15 Oct, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7868485","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":542168365,"identity":"d88f3b45-46c2-4732-b3e3-3d616e7fd334","order_by":0,"name":"Diana Immel","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABH0lEQVRIiWNgGAWjYDCCw2BSAoiZGx+AGewMBgyMDWBhAwJaGJsNwAxmQloOwFmMbRJgmpAWvuPMzx78YLCI1m1vbKvm3WHBwM/MvPFx4Y7D+fIOzBsfYNEieZjN3LCHQSJ325mDbbd5z0gwSDazFRvPPHPYcuMBtmJs1hgcZjCT4AFpuZEI1NImUb/hMI+ZNG/bYQPDBh4zCaxa2L9J/gFpuf+wrRioBSjCY/4bqsX8B1YtQDMhtjC2MUO1mDGDtMgz8Jhh877kYZ4yaRkDkF8SmyXntkH8Is17Jt3AgJmtGJvD+M4f3yb5pqIud9vxwwc/vG2rY+Bnb974mXeHtYF8e/PGD9isgTgPq+BhnOpxAfkGkrWMglEwCkbB8AQAsytd0Ovya9YAAAAASUVORK5CYII=","orcid":"","institution":"University of Oldenburg","correspondingAuthor":true,"prefix":"","firstName":"Diana","middleName":"","lastName":"Immel","suffix":""},{"id":542168366,"identity":"057f9e80-68c3-4cd2-a880-2b2608b95954","order_by":1,"name":"Elisa Mallick","email":"","orcid":"","institution":"ki:elements GmbH","correspondingAuthor":false,"prefix":"","firstName":"Elisa","middleName":"","lastName":"Mallick","suffix":""},{"id":542168367,"identity":"8b1fc84c-4dc9-404c-a25e-fad75aa56f29","order_by":2,"name":"Nicklas Linz","email":"","orcid":"","institution":"ki:elements GmbH","correspondingAuthor":false,"prefix":"","firstName":"Nicklas","middleName":"","lastName":"Linz","suffix":""},{"id":542168368,"identity":"5ec18056-5e34-4f65-bad2-06a553bcf530","order_by":3,"name":"Simon Barton","email":"","orcid":"","institution":"University of Oldenburg","correspondingAuthor":false,"prefix":"","firstName":"Simon","middleName":"","lastName":"Barton","suffix":""},{"id":542168369,"identity":"371ed514-e90a-4389-82f7-f27e83e38d61","order_by":4,"name":"Rene Hurlemann","email":"","orcid":"","institution":"University of Oldenburg","correspondingAuthor":false,"prefix":"","firstName":"Rene","middleName":"","lastName":"Hurlemann","suffix":""},{"id":542168370,"identity":"b9efea24-4c06-43c2-b420-44a3c996f2f0","order_by":5,"name":"Dirk Scheele","email":"","orcid":"","institution":"Ruhr University Bochum","correspondingAuthor":false,"prefix":"","firstName":"Dirk","middleName":"","lastName":"Scheele","suffix":""}],"badges":[],"createdAt":"2025-10-15 13:08:33","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7868485/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7868485/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-026-45965-5","type":"published","date":"2026-04-04T16:00:14+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":95662988,"identity":"99f5ebe7-8c82-4099-830c-c425e7da646a","added_by":"auto","created_at":"2025-11-11 16:38:19","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":321236,"visible":true,"origin":"","legend":"","description":"","filename":"PASoL.docx","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/35d59dccc740e3f505e554d0.docx"},{"id":95662897,"identity":"5d9306d9-2481-4115-ab50-f7f8c613c5fa","added_by":"auto","created_at":"2025-11-11 16:38:18","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7385,"visible":true,"origin":"","legend":"","description":"","filename":"e2b171494a8845a0a54cfb7544d4607c.json","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/0493c607d788a7d8019c567f.json"},{"id":95662896,"identity":"e0bb33fd-1f31-47bf-b52b-fcfc80991493","added_by":"auto","created_at":"2025-11-11 16:38:17","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":45403,"visible":true,"origin":"","legend":"","description":"","filename":"SISoL.docx","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/aa81e62507b2be1e6cf2cd29.docx"},{"id":95662820,"identity":"bec746b9-c40e-437c-86b3-dca281a36a9d","added_by":"auto","created_at":"2025-11-11 16:38:14","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":125066,"visible":true,"origin":"","legend":"","description":"","filename":"e2b171494a8845a0a54cfb7544d4607c1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/b876f7c11bed948b62bd524f.xml"},{"id":95662714,"identity":"fe682f64-87d9-4f93-a436-b4b7badd3849","added_by":"auto","created_at":"2025-11-11 16:38:02","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":209317,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/549a09c09569501a726c8a19.png"},{"id":95662839,"identity":"cbfbb91c-2b93-4a80-b44b-eb69c5ef20f7","added_by":"auto","created_at":"2025-11-11 16:38:15","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53264,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/0bcca709a221308625cd0c31.png"},{"id":95662811,"identity":"b9e8d38a-4497-48e3-b368-21a4b120627a","added_by":"auto","created_at":"2025-11-11 16:38:13","extension":"xml","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":121531,"visible":true,"origin":"","legend":"","description":"","filename":"e2b171494a8845a0a54cfb7544d4607c1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/262a06b323548bb9d3b9764f.xml"},{"id":95663120,"identity":"37a28fc2-eccd-4e96-87db-368bb49733d4","added_by":"auto","created_at":"2025-11-11 16:38:24","extension":"html","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":140997,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/da009f19e6e57ac5ff548512.html"},{"id":95662810,"identity":"73a142d1-0045-44e1-ab18-e886f6a4aaad","added_by":"auto","created_at":"2025-11-11 16:38:13","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":209317,"visible":true,"origin":"","legend":"\u003cp\u003eSpeech features extracted from the semi-guided picture description task significantly predicted loneliness in males (A) and females (B). The gray area indicates 95% confidence intervals.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/52d3657948e2afef935139fc.png"},{"id":106344104,"identity":"7ac1b3e1-3862-4c5e-aa9f-955b65bca4ec","added_by":"auto","created_at":"2026-04-07 16:12:17","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":806653,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/ae6076bf-a81e-4a3a-8d41-15b825491c61.pdf"},{"id":95662816,"identity":"dd5bdf78-d6c5-4304-a746-4f83f701fa59","added_by":"auto","created_at":"2025-11-11 16:38:14","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":45403,"visible":true,"origin":"","legend":"","description":"","filename":"SISoL.docx","url":"https://assets-eu.researchsquare.com/files/rs-7868485/v1/e9e1e4241298c0f50e36f1a8.docx"}],"financialInterests":"Competing interest reported. Elisa Mallick is employed by the company ki:elements, which developed the application for the speech-based assessment and extracted the speech features. Nicklas Linz owns shares in the ki:elements company. Dirk Scheele, Simon Barton, Rene Hurlemann and Diana Immel have nothing to disclose.","formattedTitle":"The sound of loneliness: Prediction of perceived social isolation using automatic speech analysis","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHuman beings are social by nature and are intrinsically motivated to form and maintain interpersonal relationships. When a person's need to belong is not consistently met, feelings of loneliness develop. Loneliness has a cross-age prevalence rate of up to 33% and has detrimental effects on mental and physical health. It is associated with an increased risk of psychiatric disorders such as depression [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e], psychosis [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], and social anxiety [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Importantly, it is also associated with suicidality [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e] and higher all-cause mortality [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Multiple lines of evidence indicate that loneliness is associated with negative cognitive biases [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. For instance, implicit hypervigilance to social threats, increased anticipation of rejection, and negative attributional styles can result in higher expectations of negative social interactions. At the neural level, reduced reactivity and functional connectivity of the anterior insula may reflect impaired integration of trust-related information in lonely individuals [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Despite the phenotypic overlap, the neural mechanisms associated with loneliness appear to be distinct from those associated with social anxiety [10]. Importantly, lonely individuals tend to elicit behaviors from interaction partners that lead to the confirmation of their negative expectations, resulting in a self-fulfilling prophecy [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. This theoretical framework also helps to explain why subtle alterations in speech may emerge as markers of loneliness because communicative behavior reflects and reinforces social expectations. Consequently, lonely individuals prefer greater social distance in interactions with strangers and benefit less from positive social interactions [9, 12]. Meta-analytical evidence also indicates a small negative association between loneliness and prosociality [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eAlthough several studies support the notion that lonely individuals have impaired social interactions, the underlying behavioral mechanisms remain elusive. Preliminary evidence suggests that loneliness is associated with emotion-specific impairments in the recognition of vocal expressions [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. However, given the bidirectional nature of social interactions, it is conceivable that both altered speech recognition and production may result in dysfunctional social interactions. Previously, we found that the hypothalamic peptide oxytocin facilitates communicative reciprocity by enhancing the salience of vocal expressions [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Additionally, we found that interaction-induced oxytocin release is significantly impaired in lonely individuals [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. These oxytocin-related deficits provide a theoretical basis for why loneliness may be detectable through paralinguistic speech features. Consistent with this, self-reports of communicative competence negatively correlate with loneliness [15].\u003c/p\u003e\u003cp\u003eNotably, machine learning experiments are increasingly being used to identify speech patterns in patients with mental disorders. Automated speech analysis has been employed to detect apathy in older adults with cognitive impairments and to classify patients with affective disorders [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e16\u003c/span\u003e], posttraumatic stress disorder [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e17\u003c/span\u003e], and suicidal ideation [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. A recent study found that machine learning models based on natural language processing of manually transcribed qualitative interviews can be used to predict loneliness in older adults [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. However, given the emotion-specific effects of loneliness on speech production, it is still unclear whether changes in speech features associated with loneliness are more pronounced in emotional content and in older individuals, and whether a dyadic setting is required.\u003c/p\u003e\u003cp\u003eTo this end, we used the revised UCLA Loneliness Questionnaire (UCLA) [20, 21] to assess loneliness in 96 healthy participants (53 women, mean age 30.85 years; 43 men, mean age 31.37 years). We recorded the participants\u0026rsquo; speech during three tasks. They described the Cookie Theft picture from the Boston Diagnostic Aphasia Examination [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e22\u003c/span\u003e] and were asked to talk about positive and negative life events [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. We used Random Forest regression models to predict the UCLA score from extracted acoustic speech features. Because speech features naturally vary by sex, we conducted the analyses separately for women and men. Due to the phenotypic overlap between loneliness, depression, and social anxiety, we also assessed depression and social anxiety as control variables. We hypothesized that loneliness would be associated with significantly altered paralinguistic markers relating to the prosodic, formant, source, and temporal qualities of speech [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eWe observed average loneliness scores of 41.72 in women and 38.02 in men (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) which are comparable to the normative values reported for the UCLA Loneliness Scale in college students [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. There were no significant differences in loneliness, depression, or social anxiety between women and men (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). In both genders, loneliness significantly correlated with depression (women: \u003cem\u003er\u003c/em\u003e\u003csub\u003e(53)\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.59, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001; men: \u003cem\u003er\u003c/em\u003e\u003csub\u003e(43)\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.54, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001) and social anxiety (women: \u003cem\u003er\u003c/em\u003e\u003csub\u003e(53)\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.34, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.005; men: \u003cem\u003er\u003c/em\u003e\u003csub\u003e(43)\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.32, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.05). All questionnaires showed high internal consistency (Cronbach\u0026rsquo;s Alpha\u0026thinsp;\u0026gt;\u0026thinsp;0.9; see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSample description\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFemales\u003c/p\u003e\u003cp\u003e\u003cem\u003eM\u003c/em\u003e (\u003cem\u003eSD\u003c/em\u003e)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMales\u003c/p\u003e\u003cp\u003e\u003cem\u003eM\u003c/em\u003e (\u003cem\u003eSD\u003c/em\u003e)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eU\u003c/em\u003e (96)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eCronbachs Alpha\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge (years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e30.85\u003c/p\u003e\u003cp\u003e(10.90)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e31.37\u003c/p\u003e\u003cp\u003e(9.97)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1246.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEducation (years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e16.28\u003c/p\u003e\u003cp\u003e(2.82)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e16.53\u003c/p\u003e\u003cp\u003e(2.42)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1182.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLoneliness\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e41.72\u003c/p\u003e\u003cp\u003e(14.71)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e38.02\u003c/p\u003e\u003cp\u003e(13.34)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e934\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e.96\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDepression\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.13\u003c/p\u003e\u003cp\u003e(6.29)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6.98\u003c/p\u003e\u003cp\u003e(7.88)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1186.5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e.90\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSocial Anxiety\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e15.85\u003c/p\u003e\u003cp\u003e(10.86)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e12.14\u003c/p\u003e\u003cp\u003e(11.15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e862.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e.95\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eImportantly, the set of speech features extracted from the picture description task by machine learning models significantly predicted loneliness scores of both women and men (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The model explained 6% and 16% of the variance in women and men, respectively (see Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e\u003cb\u003e).\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAs expected, depression and social anxiety also predicted loneliness in the machine learning analyses. For women, a combined model that included speech features from the picture description, depression, and social anxiety scores as predictors explained more variance (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;.21, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.01) than reduced models that included either speech features from the picture description (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;.06, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.04) or depression and social anxiety scores (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;.16, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.02). However, in men, the combined model with speech features, depression, and social anxiety scores (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;.16, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.03) resulted in a worse fit than the reduced model without the picture description speech features (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;.28, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.01) (see Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eResults of machine learning experiments for loneliness (picture description)\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"10\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colspan=\"5\" nameend=\"c6\" namest=\"c2\"\u003e\u003cp\u003eFemales\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"4\" nameend=\"c10\" namest=\"c7\"\u003e\u003cp\u003eMales\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFeatures set\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003eRandom Forest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"3\" nameend=\"c6\" namest=\"c4\"\u003e\u003cp\u003eRandomised Baseline\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c8\" namest=\"c7\"\u003e\u003cp\u003eRandom Forest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c10\" namest=\"c9\"\u003e\u003cp\u003eRandomised Baseline\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMAE\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eM\u003c/em\u003e (\u003cem\u003eSD\u003c/em\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e\u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eMAE\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e\u003cem\u003eR\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e\u003cem\u003eM\u003c/em\u003e (\u003cem\u003eSD\u003c/em\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePicture Description\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e13.39 (0.95)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e.04\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e10.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e12.50 (0.92)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e\u003cb\u003e.02\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDepression\u0026thinsp;+\u0026thinsp;Social Anxiety\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e14.35 (1.48)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e.02\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e8.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e13.10 (1.47)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;.01\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePicture Description\u0026thinsp;+\u0026thinsp;Depression\u0026thinsp;+\u0026thinsp;Social Anxiety\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e10.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e13.36 (0.95)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;.01\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e10.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e12.51 (0.91)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003e\u003cb\u003e.03\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBaseline\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e13.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e12.08\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"10\"\u003e\u003cem\u003eNotes.\u003c/em\u003e MAE: Mean absolute error of random forest regression.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eNot surprisingly, no single speech feature emerged as a strong predictor of loneliness. For picture description, significant correlations between extracted speech features and loneliness were evident in the temporal and source categories (see Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Specifically, higher loneliness was significantly associated with a lower speech to non-speech ratio in women. Similarly, a higher kurtosis value in the amplitude distribution of the signal from the picture description task was associated with greater loneliness. This suggests that volume intensity was distributed more irregularly among lonely women. In men, greater loneliness significantly correlated with fewer pauses between syllables and a shorter phonation time. Additionally, men with higher loneliness scores had a lower sound-to-noise ratio, reflecting poorer voice quality, and a higher mean pitch.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eTOP 5 highest spearman rank partial correlations between speech features and loneliness, for the picture description\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colspan=\"5\" nameend=\"c6\" namest=\"c2\"\u003e\u003cp\u003eFemales\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSpeech features\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSpeech ratio\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAmplitude kurtosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003ePeak frequency\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eAmplitude mean absolute value\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eMean power\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCoefficient\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026minus;\u0026thinsp;.29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026minus;\u0026thinsp;.22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026minus;\u0026thinsp;.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e.13\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e.03\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e.04\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e.14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e.34\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colspan=\"5\" nameend=\"c6\" namest=\"c2\"\u003e\u003cp\u003e\u003cb\u003eMales\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSpeech features\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNumber of pauses\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTotal phonation time\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSound to noise ratio\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eHarmonics to noise ratio\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePower spectrum ratio\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCoefficient\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026minus;\u0026thinsp;.44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026minus;\u0026thinsp;.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026minus;\u0026thinsp;.37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026minus;\u0026thinsp;.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026minus;\u0026thinsp;.21\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;.01\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;.01\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e.02\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e.10\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e.17\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe set of speech features extracted from the free speech emotional storytelling tasks did not significantly predict loneliness in either positive or negative storytelling for women or men. As expected, the combined models (negative/positive storytelling\u0026thinsp;+\u0026thinsp;depression\u0026thinsp;+\u0026thinsp;social anxiety) explained more variance than the reduced models that examined only storytelling for both genders. However, the combined model only became significant (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;.09, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.04) for negative storytelling among women (see \u003cb\u003eTables S1, S2\u003c/b\u003e).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn the present study, we examined whether loneliness is reflected in speech features in a heterogeneous sample of young healthy adults. Using a machine learning-based statistical approach, we found that paralinguistic markers extracted from a semi-guided picture description task significantly predicted loneliness in both women and men. Speech features from the temporal and source categories appear to be particularly relevant to this association. Interestingly, a model that included both speech features and depression and social anxiety scores enabled a better prediction than a model only with psychiatric symptoms in women, but not men. However, extraction of speech features from positive and negative free storytelling did not significantly predict loneliness.\u003c/p\u003e\u003cp\u003eLoneliness can affect social interactions in numerous ways. For instance, highly lonely individuals prefer greater distance from an unfamiliar interaction partner [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e], exhibit altered gaze processing [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e], and increased gaze towards their conversation partners' faces [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Sleep-deprived participants have been rated as significantly lonelier and less desirable to interact with [27]. Furthermore, blinded experimenters were able to identify whether they were interacting with a lonely or non-lonely individual [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. A previous study found that loneliness could be predicted from the content of transcribed speech using natural language processing in older adults [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. However, our findings suggest that loneliness is also reflected in paralinguistic markers. This highlights an innovative shift from \u0026ldquo;what is said\u0026rdquo; to \u0026ldquo;how it is said.\u0026rdquo; Consistent with the multifaceted nature of loneliness, the present proof-of-concept study builds on previous approaches of natural language processing by demonstrating that alterations in speech related to loneliness are not limited to linguistic content, but can also be detected in paralinguistic domains (e.g., temporal, source-related, spectral or prosodic speech categories).\u003c/p\u003e\u003cp\u003eInterestingly, speech features extracted from emotional storytelling did not significantly predict loneliness. Arousal-induced speech changes may have obscured loneliness-specific markers. Another possibility is that the Cookie Theft scenario triggered stronger social-cognitive processing because participants had to infer intentions, roles, and relationships between characters. These demands may directly activate biases related to loneliness in attention and interpretation, which could be reflected in paralinguistic speech markers. In contrast, free storytelling about personal life events may rely more on autobiographical memory and emotional arousal. These processes could overshadow subtle, loneliness-related alterations in speech.\u003c/p\u003e\u003cp\u003ePreviously, a reduced oxytocinergic response to semi-guided social interactions was observed in individuals with high loneliness [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Reduced oxytocin release may impair the transmission of emotional information in social settings because exogenous (e.g. nasally administered) oxytocin enhanced facial and vocal expression of fear and happiness [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. These observations highlight that the detectability of loneliness-related speech markers is likely task- and context-dependent. Further research is also needed to investigate whether changes in speech features are related to endocrine function.\u003c/p\u003e\u003cp\u003eAlthough loneliness is an important risk factor for depression and anxiety, accumulating evidence suggests that it should be considered a distinct construct. In a prospective longitudinal study, loneliness predicted subsequent changes in depressive symptomatology, but not vice versa [28]. Similarly, loneliness exhibits a unique neural profile during cognitive control tasks in patients with major depressive disorder and in healthy controls [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Additionally, evidence suggests that lonely and non-lonely individuals experience equal subjective valence when engaging in social situations, as well as exhibit comparable amygdala responses to social decisions and striatal responses to positive social feedback [10]. This pattern of responses stands in stark contrast to the findings for social anxiety [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. In the present study, loneliness significantly correlated with depression and social anxiety in both women and men. Interestingly, the combined predictive model, which included speech features from the Picture Description Task, as well as depression and social anxiety scores, provided a better model fit for females. For males, however, this model showed a poorer fit than the reduced models, a tendency also evident in the storytelling tasks. These results suggest that loneliness may follow gender-specific pathways. Prior studies support such differences. Specifically, loneliness has been associated with a more pronounced within-network coupling of the default network in men than in women [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. An interaction between loneliness and gender was also found following an experimental trauma paradigm: more intrusions were reported by lonely men, but not by lonely women [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. This effect was accompanied by reduced amygdala habituation to repeated fearful faces and amygdala hyperreactivity during fear conditioning in lonely men. Our results contribute to the existing literature by suggesting that the prediction of loneliness through speech may be more strongly moderated by comorbid symptoms in men than in women. While the driving mechanisms remain unclear, emphasizing gender as a potential moderator is an important direction for future hypothesis-driven research on sex-specific pathways of social communication.\u003c/p\u003e\u003cp\u003eThere are several limitations to the current study. First, we recruited healthy individuals with varying levels of loneliness, so it is unclear whether our findings can be generalized to patients with depression or anxiety disorders. Additionally, chronic loneliness is a relatively stable construct with trait-like properties [33]. However, it is likely that the adverse health consequences of loneliness depend on its chronicity [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Even brief periods of social isolation can lead to decreased energy levels and increased feelings of fatigue [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e35\u003c/span\u003e], but situational loneliness seems to drive people toward reconnection, while chronic loneliness seems to drive people away from it [12]. We assessed trait-like loneliness using the established UCLA Loneliness Scale, which does not allow conclusions about the chronicity of the perceived social isolation.\u003c/p\u003e\u003cp\u003e Taken together, these findings provide the first evidence that loneliness can be predicted by paralinguistic markers that are automatically extracted from semi-guided speech. This mechanism may explain why loneliness can be perceived by others and shed light on a pathway by which loneliness may hinder positive interactions, thereby propagating the maintenance of chronic loneliness. Future research should test these approaches in larger, more diverse samples, including clinical populations, and adopt longitudinal designs that capture loneliness dynamics over time. Furthermore, incorporating speech-based assessments into ecologically valid settings, such as everyday social interactions or digital health platforms, could substantially increase their translational potential.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e4.1 Participants\u003c/h2\u003e\u003cp\u003eEligibility for the study included the following requirements: Participants had to be between 18 and 65 years of age, speak sufficient German, and not have a psychiatric diagnosis or be taking psychiatric medication. Two sources were used to recruit participants. First, healthy participants were recruited through online advertisements and public notices. Second, pre-stratified healthy participants in a group therapy intervention aimed at reducing loneliness were asked to participate in the study before starting the therapy intervention. A total of 105 participants were included in the study. Nine participants were excluded from the analyses due to missing voice recordings or other missing data. The final sample consisted of \u003cem\u003eN\u003c/em\u003e\u0026thinsp;=\u0026thinsp;96 people (53 women and 43 men). The mean age was 30.85 years (\u0026plusmn;\u0026thinsp;\u003cem\u003eSD\u003c/em\u003e: 10.90) for women and 31.37 years (\u0026plusmn;\u0026thinsp;\u003cem\u003eSD\u003c/em\u003e: 9.97) for men. The study was approved by the Ethics Committee of the University Hospital of Bonn and was conducted according to the principles of the Declaration of Helsinki. The study and data analyses were pre-registered (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://osf.io/buqrj/\u003c/span\u003e\u003cspan address=\"https://osf.io/buqrj/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Participants were enrolled after providing written informed consent and received monetary compensation at the end of the study.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e4.2 Study Tasks\u003c/h2\u003e\u003cp\u003eThe \u0026ldquo;Cookie Theft\u0026rdquo; picture from the Boston Diagnostic Aphasia Examination is a well-established method for assessing the expressive language skills of children and adults [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. One feature of the task is that it elicits mental state language [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. The picture depicts a familiar domestic scene that requires making assumptions about the mental states of the characters. For instance, the mother is daydreaming and therefore does not notice her children climbing on a stool that is about to fall while they scramble for biscuits. In the free emotional storytelling task, the participants were asked to talk about a negative and a positive event in their lives [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e4.3 Questionnaires\u003c/h2\u003e\u003cp\u003eIn addition to the speech assessment, clinical measures were also collected as part of the investigation. The Becks Depression Inventory (BDI-II) is a psychological self-report instrument (21 items with a 4-point Likert scale) for assessing the severity of depression in adolescents and adults ranging from 0 to 63 [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. The Liebowitz Social Anxiety Scale (LSAS) (50 items with a 4-point Likert scale) is a questionnaire with a range from 0 to 72 used for the diagnosis of social anxiety disorder [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. Trait-like loneliness was measured using a validated German version of the Revised UCLA Loneliness Scale (UCLA) [40], which is a 20-item, 4-point Likert scale with scores ranging from 20 to 80. Numerous validation studies have established loneliness as a distinct psychological construct [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], [41]. Psychometric test properties, such as retest reliability and internal consistency, are considered satisfactory [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e42\u003c/span\u003e, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e4.4 Procedure\u003c/h2\u003e\u003cp\u003eAll participants attended one 125-minute study session. The objective of the study and the study procedure were explained. The inclusion and exclusion criteria were explicitly assessed, and written consent was requested before the assessment began. All participants were screened using the Mini-International Neuropsychiatric Interview (MINI) [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. Then, psychometric questionnaires were administered using Qualtrics software (Provo, USA). Then, the speech assessment was administered on an Apple iPad tablet performed by the ∆elta Clinical app [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. This study was part of a larger study, the results of which are described elsewhere. The speech assessment took approximately five minutes per task and was conducted with an experimenter present. During the speech tasks, the tablet recorded the participants' speech features.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e4.5 Data Analysis\u003c/h2\u003e\u003cp\u003eThe speech data consist of various speech features (see \u003cb\u003eTable S8\u003c/b\u003e, that were automatically extracted from the audio signal by the iOS app ∆elta Clinical [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e46\u003c/span\u003e]). These features were extracted separately for picture description, positive storytelling, and negative storytelling. They are grouped into four main categories:\u003c/p\u003e\u003cp\u003e\u003col\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003e\u003cem\u003eTemporal features\u003c/em\u003e indicate the general rate of speech and measure the proportion of speech (e.g., length and connection of speech segments and the pauses between them). These features reflect the effectiveness of speech production and overlap with prosodic speech characteristics, in the form of fluency and rhythm.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003e\u003cem\u003eProsodic features\u003c/em\u003e refer to the long-term dynamics of perceived intonation and speech rhythm. These features demonstrate the overall speech melody adapted to a given situation, thereby indicating prosodic competence in terms of appropriate of speech intonation [47, 48]. Prosodic features also measure changes in an individual's speaking style (e.g., perceived intonation or pitch).\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003e\u003cem\u003eSpectral features\u003c/em\u003e represent the relationship between articulatory movements and changes in vocal tract shape. These features include spectral flow, energy, slope and flatness [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]. Spectral features measure the airborne noise caused by the speech signal and the power of the strongest frequency relative to all others, such as background noise, which can be filtered out to improve speech analysis.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003e\u003cem\u003eSource features\u003c/em\u003e are important markers of voice quality. They represent the auditory perceptibility of changes in vocal fold vibration and vocal tract shape, outside of pitch, loudness, and phonetics. Source features frequently record information about laryngeal qualities, such as breathing, creaking, hardness, and phonation type [50].\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003c/ol\u003e\u003c/p\u003e\u003cp\u003eDue to noise (e.g., background noise) in the processing, the speech features espinola zero crossing metric, mean F0, and average amplitude change demonstrated zero values and were excluded from the analysis.\u003c/p\u003e\u003cdiv id=\"Sec10\" class=\"Section3\"\u003e\u003ch2\u003e4.5.1 Statistical Analysis\u003c/h2\u003e\u003cp\u003eThe database contains 78 speech features and three clinical scores (UCLA, BDI-II and LSAS) from 105 individuals. To account for gender-specific differences in speech characteristics between women and men [51], the data were analyzed separately by gender. QQ plots and Shapiro-Wilk tests revealed that the UCLA, BDI-II and LSAS scores did not follow a normal distribution. The total LSAS score was obtained by summing the anxiety and avoidance subscores. During the data analysis process, it was decided to deviate from the originally planned registration. The variable \"loneliness\" was used as a continuous variable to mitigate the loss of information. Consequently, the area under the curve calculation, which is used to make predictions, was not conducted. Statistical analysis was performed with Rstudio (version 1.4.1103). Spearman rank correlations were calculated between UCLA, BDI-II and LSAS scores and speech characteristics for both sexes. Demographic and psychological variables were compared between sexes with Mann-Whitney-U-tests. Internal consistency of the three clinical scores (UCLA, BDI-II, and LSAS) was calculated (IBM SPSS Statistics (Version 30)) using Cronbach's alpha, a statistical method for measuring internal consistency in scales or inventories.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section3\"\u003e\u003ch2\u003e4.5.2 Machine learning experiments\u003c/h2\u003e\u003cp\u003eRandom forest regression models were used to predict the UCLA score based on acoustic speech features extracted from speech tasks, as well as the BDI-II and LSAS scores. The features were normalized using a standard scaler. The models were trained using leave-one-out cross-validation and grid search for hyperparameter tuning. Mean absolute error (MAE) and R-squared are reported as performance measures. The model results were compared with the baseline MAE obtained by predicting the population mean. To calculate the statistical significance of the regression models' performance, a randomized baseline was used, consisting of training an extra tree model several times with the labels permuted each time.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section3\"\u003e\u003ch2\u003e4.5.3 Power Analyses\u003c/h2\u003e\u003cp\u003eTo date, no study has examined the relationship between loneliness and automatically extracted speech features. Therefore, an a priori power analysis was conducted for this project using G*Power 3. This analysis was based on the effect size obtained in a previous study that examined the effects of loneliness on affective responsiveness to a positive social interactions [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. The results showed that the positive mood change induced by an interaction was significantly reduced in participants with high loneliness (\u003cem\u003er\u003c/em\u003e(79)\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;.25, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.03). To reliably replicate this effect of loneliness (with α\u0026thinsp;=\u0026thinsp;.05, and power\u0026thinsp;=\u0026thinsp;.80, one-tailed \u003cem\u003et\u003c/em\u003e-test), at least 95 participants must be tested. To account for possible dropouts, the plan was to test at least 100 participants (50 women).\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003eConflict of interest statement\u003c/p\u003e\n\u003cp\u003eElisa Mallick is employed by the company ki:elements, which developed the application for the speech-based assessment and extracted the speech features. Nicklas Linz owns shares in the ki:elements company. Dirk Scheele, Simon Barton, Rene Hurlemann and Diana Immel have nothing to disclose.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFunding sources\u003c/p\u003e\n\u003cp\u003eR.H. and D.S. were supported by a grant from the German Research Foundation (DFG) (HU 1302/18-1 and SCHE 1913/7-1).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAuthors\u0026apos; contributions\u003c/p\u003e\n\u003cp\u003eD.I. and D.S. designed the experiment; D.I. performed the experiments; D.I., E.M., S.B. and N.L. analysed the data. All authors drafted the manuscript. All authors read and approved the current version of the manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Data availability statement\u003c/p\u003e\n\u003cp\u003eThe data will be provided upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eCacioppo, J. T., Hughes, M. E., Waite, L. J., Hawkley, L. C. \u0026amp; Thisted, R. A. Loneliness as a specific risk factor for depressive symptoms: Cross-sectional and longitudinal analyses. \u003cem\u003ePsychol. Aging\u003c/em\u003e\u003cstrong\u003e21\u003c/strong\u003e, 140\u0026ndash;151 (2006). doi: 10.1037/0882-7974.21.1.140\u003c/li\u003e\n\u003cli\u003eVelthorst, E. et al. The impact of loneliness and social relationship dissatisfaction on clinical and functional outcomes in Dutch mental health service users. \u003cem\u003ePsychiatry Res.\u003c/em\u003e\u003cstrong\u003e342\u003c/strong\u003e, 116242 (2024). doi: 10.1016/j.psychres.2024.116242\u003c/li\u003e\n\u003cli\u003eMaes, M. et al. Loneliness and social anxiety across childhood and adolescence: Multilevel meta-analyses of cross-sectional and longitudinal associations. \u003cem\u003eDev. Psychol.\u003c/em\u003e\u003cstrong\u003e55\u003c/strong\u003e, 1548\u0026ndash;1565 (2019). doi: 10.1037/dev0000745\u003c/li\u003e\n\u003cli\u003eInstitute of Medicine. \u003cem\u003eReducing Suicide: A National Imperative\u003c/em\u003e. (National Academies Press, 2002). doi: 10.17226/10398\u003c/li\u003e\n\u003cli\u003eHolt-Lunstad, J., Smith, T. B., Baker, M., Harris, T. \u0026amp; Stephenson, D. Loneliness and social isolation as risk factors for mortality: a meta-analytic review. \u003cem\u003ePerspect. Psychol. Sci.\u003c/em\u003e\u003cstrong\u003e10\u003c/strong\u003e, 227\u0026ndash;237 (2015). doi: 10.1177/1745691614568352\u003c/li\u003e\n\u003cli\u003eHawkley, L. C. \u0026amp; Cacioppo, J. T. Loneliness matters: A theoretical and empirical review of consequences and mechanisms. \u003cem\u003eAnn. Behav. Med.\u003c/em\u003e\u003cstrong\u003e40\u003c/strong\u003e, 218\u0026ndash;227 (2010). doi: 10.1007/s12160-010-9210-8\u003c/li\u003e\n\u003cli\u003eMalon, M., Gajos, K., Rajchert, J., Holt-Lunstad, J. \u0026amp; Okruszek, Ł. Lonely and self-centered? A meta-analysis of the link between prosociality and loneliness. \u003cem\u003ePers. Soc. Psychol. Bull.\u003c/em\u003e (2024). doi: 10.1177/01461672241295263\u003c/li\u003e\n\u003cli\u003eSpithoven, A. W. M., Bijttebier, P. \u0026amp; Goossens, L. It is all in their mind: A review on information processing bias in lonely individuals. \u003cem\u003eClin. Psychol. Rev.\u003c/em\u003e\u003cstrong\u003e58\u003c/strong\u003e, 97\u0026ndash;114 (2017). doi: 10.1016/j.cpr.2017.10.003\u003c/li\u003e\n\u003cli\u003eLieberz, J. et al. Loneliness and the social brain: How perceived social isolation impairs human interactions. bioRxiv (2021). doi: 10.1002/advs.202102076\u003c/li\u003e\n\u003cli\u003eCacioppo, J. T., Norris, C. J., Decety, J., Monteleone, G. \u0026amp; Nusbaum, H. In the eye of the beholder: Individual differences in perceived social isolation predict regional brain activation to social stimuli. \u003cem\u003eJ. Cogn. Neurosci.\u003c/em\u003e\u003cstrong\u003e21\u003c/strong\u003e, 83\u0026ndash;92 (2009). doi: 10.1162/jocn.2008.21007\u003c/li\u003e\n\u003cli\u003eSaporta, N. et al. Opposing association of situational and chronic loneliness with interpersonal distance. \u003cem\u003eBrain Sci.\u003c/em\u003e\u003cstrong\u003e11\u003c/strong\u003e, 1135 (2021). doi: 10.3390/brainsci11081135\u003c/li\u003e\n\u003cli\u003eMorningstar, M., Nowland, R., Dirks, M. A. \u0026amp; Qualter, P. Loneliness and the recognition of vocal socioemotional expressions in adolescence. \u003cem\u003eCognition Emot.\u003c/em\u003e\u003cstrong\u003e34\u003c/strong\u003e, 970\u0026ndash;976 (2020). doi: 10.1080/02699931.2019.1688602\u003c/li\u003e\n\u003cli\u003eSpengler, F. B. et al. Oxytocin facilitates reciprocity in social communication. \u003cem\u003eSoc. Cogn. Affect. Neurosci.\u003c/em\u003e\u003cstrong\u003e12\u003c/strong\u003e, 1325\u0026ndash;1333 (2017). doi: 10.1093/scan/nsx056\u003c/li\u003e\n\u003cli\u003eZakahi, W. R. \u0026amp; Duran, R. L. Loneliness, communicative competence, and communication apprehension: Extension and replication. \u003cem\u003eCommun. Q.\u003c/em\u003e\u003cstrong\u003e33\u003c/strong\u003e, 50\u0026ndash;60 (1985). doi: 10.1080/01463378509369696\u003c/li\u003e\n\u003cli\u003eFlanagan, O., Chan, A., Roop, P. \u0026amp; Sundram, F. Using acoustic speech patterns from smartphones to investigate mood disorders: Scoping review. \u003cem\u003eJMIR Mhealth Uhealth\u003c/em\u003e\u003cstrong\u003e9\u003c/strong\u003e, e24352 (2021). doi: 10.2196/24352\u003c/li\u003e\n\u003cli\u003eMarmar, C. R. et al. Speech-based markers for posttraumatic stress disorder in US veterans. \u003cem\u003eDepress. Anxiety\u003c/em\u003e\u003cstrong\u003e36\u003c/strong\u003e, 607\u0026ndash;616 (2019). doi: 10.1002/da.22946\u003c/li\u003e\n\u003cli\u003eBelouali, A. et al. Acoustic and language analysis of speech for suicidal ideation among US veterans. \u003cem\u003eBioData Mining\u003c/em\u003e\u003cstrong\u003e14\u003c/strong\u003e, 11 (2021). doi: 10.1186/s13040-021-00245-0\u003c/li\u003e\n\u003cli\u003eBadal, V. D. et al. Prediction of loneliness in older adults using natural language processing: Exploring sex differences in speech. \u003cem\u003eAm. J. Geriatr. Psychiatry\u003c/em\u003e\u003cstrong\u003e29\u003c/strong\u003e, 853\u0026ndash;866 (2021). doi: 10.1016/j.jagp.2021.02.017\u003c/li\u003e\n\u003cli\u003eRussell, D., Peplau, L. A. \u0026amp; Cutrona, C. E. The revised UCLA Loneliness Scale: Concurrent and discriminant validity evidence. \u003cem\u003eJ. Pers. Soc. Psychol.\u003c/em\u003e\u003cstrong\u003e39\u003c/strong\u003e, 472\u0026ndash;480 (1980). doi: 10.1037/0022-3514.39.3.472\u003c/li\u003e\n\u003cli\u003eSchwab, R. Zur Messung und klinisch-psychologischen Bedeutung von Einsamkeit: Eine deutschsprachige Version der \u0026bdquo;Rev. UCLA Loneliness Scale\u0026ldquo; von Russell, Peplau \u0026amp; Cutrona (1980). In \u003cem\u003eBericht \u0026uuml;ber den 13. Kongre\u0026szlig; f\u0026uuml;r Angewandte Psychologie\u003c/em\u003e, Bd 2, 245\u0026ndash;267 (Bonn, 1985).\u003c/li\u003e\n\u003cli\u003eGiles, E., Patterson, K. \u0026amp; Hodges, J. R. Performance on the Boston Cookie Theft picture description task in patients with early dementia of the Alzheimer\u0026rsquo;s type: Missing information. \u003cem\u003eAphasiology\u003c/em\u003e\u003cstrong\u003e10\u003c/strong\u003e, 395\u0026ndash;408 (1996). doi: 10.1080/02687039608248460\u003c/li\u003e\n\u003cli\u003eK\u0026ouml;nig, A. et al. Detecting apathy in older adults with cognitive disorders using automatic speech analysis. \u003cem\u003eJ. Alzheimers Dis.\u003c/em\u003e\u003cstrong\u003e69\u003c/strong\u003e, 1183\u0026ndash;1193 (2019). doi: 10.3233/JAD-180947\u003c/li\u003e\n\u003cli\u003eBell, V. et al. Do loneliness and social exclusion breed paranoia? An experience sampling investigation across the psychosis continuum. \u003cem\u003eSchizophr. Res. Cogn.\u003c/em\u003e\u003cstrong\u003e33\u003c/strong\u003e, 100282 (2023). doi: 10.1016/j.scog.2023.100282\u003c/li\u003e\n\u003cli\u003eKanai, R. et al. Brain structure links loneliness to social perception. \u003cem\u003eCurr. Biol.\u003c/em\u003e\u003cstrong\u003e22\u003c/strong\u003e, 1975\u0026ndash;1979 (2012). doi: 10.1016/j.cub.2012.08.030\u003c/li\u003e\n\u003cli\u003eLodder, G. M. A., Scholte, R. H. J., Goossens, L., Engels, R. C. M. E. \u0026amp; Verhagen, M. Loneliness and the social monitoring system: Emotion recognition and eye gaze in a real-life conversation. \u003cem\u003eBr. J. Psychol.\u003c/em\u003e\u003cstrong\u003e107\u003c/strong\u003e, 135\u0026ndash;153 (2016). doi: 10.1111/bjop.12114\u003c/li\u003e\n\u003cli\u003eBen Simon, E. \u0026amp; Walker, M. P. Sleep loss causes social withdrawal and loneliness. \u003cem\u003eNat. Commun.\u003c/em\u003e\u003cstrong\u003e9\u003c/strong\u003e, 3146 (2018). doi: 10.1038/s41467-018-05634-1\u003c/li\u003e\n\u003cli\u003eCacioppo, J. T., Hawkley, L. C. \u0026amp; Thisted, R. A. Perceived social isolation makes me sad: 5-year cross-lagged analyses of loneliness and depressive symptomatology in the Chicago Health, Aging, and Social Relations Study. \u003cem\u003ePsychol. Aging\u003c/em\u003e\u003cstrong\u003e25\u003c/strong\u003e, 453\u0026ndash;463 (2010). doi: 10.1037/a0017216\u003c/li\u003e\n\u003cli\u003eShao, R. et al. Loneliness and depression dissociated on parietal-centered networks in cognitive and resting states. \u003cem\u003ePsychol. Med.\u003c/em\u003e\u003cstrong\u003e50\u003c/strong\u003e, 2691\u0026ndash;2701 (2020). doi: 10.1017/S0033291719003096\u003c/li\u003e\n\u003cli\u003eSchultz, J. et al. A human subcortical network underlying social avoidance revealed by risky economic choices. \u003cem\u003eeLife\u003c/em\u003e\u003cstrong\u003e8\u003c/strong\u003e, e45249 (2019). doi: 10.7554/eLife.45249\u003c/li\u003e\n\u003cli\u003eSpreng, R. N. et al. The default network of the human brain is associated with perceived social isolation. \u003cem\u003eNat. Commun.\u003c/em\u003e\u003cstrong\u003e11\u003c/strong\u003e, 6393 (2020). doi: 10.1038/s41467-020-20243-0\u003c/li\u003e\n\u003cli\u003eMorr, M. et al. Lonely in the dark: Trauma memory and sex-specific dysregulation of amygdala reactivity to fear signals. \u003cem\u003eAdv. Sci.\u003c/em\u003e\u003cstrong\u003e9\u003c/strong\u003e, 2105336 (2022). doi: 10.1002/advs.202105336\u003c/li\u003e\n\u003cli\u003eMund, M., Freuding, M. M., M\u0026ouml;bius, K., Horn, N. \u0026amp; Neyer, F. J. The stability and change of loneliness across the life span: A meta-analysis of longitudinal studies. \u003cem\u003ePers. Soc. Psychol. Rev.\u003c/em\u003e\u003cstrong\u003e24\u003c/strong\u003e, 24\u0026ndash;52 (2020). doi: 10.1177/1088868320902698\u003c/li\u003e\n\u003cli\u003eHaihambo, N., Layiwola, D.-M., Blank, H., Hurlemann, R. \u0026amp; Scheele, D. Loneliness and social conformity: A predictive processing perspective. \u003cem\u003eAnn. N. Y. Acad. Sci.\u003c/em\u003e\u003cstrong\u003e1547\u003c/strong\u003e, 5\u0026ndash;17 (2025). doi: 10.1111/nyas.14853\u003c/li\u003e\n\u003cli\u003eStijovic, A. et al. Homeostatic regulation of energetic arousal during acute social isolation: Evidence from the lab and the field. \u003cem\u003ePsychol. Sci.\u003c/em\u003e\u003cstrong\u003e34\u003c/strong\u003e, 537\u0026ndash;551 (2023). doi: 10.1177/09567976221138099\u003c/li\u003e\n\u003cli\u003eGoodglass, H., Kaplan, E. \u0026amp; Weintraub, S. \u003cem\u003eBDAE: The Boston Diagnostic Aphasia Examination\u003c/em\u003e. Lippincott Williams \u0026amp; Wilkins, Philadelphia, PA (2001).\u003c/li\u003e\n\u003cli\u003eCummings, L. Describing the Cookie Theft picture: Sources of breakdown in Alzheimer\u0026rsquo;s dementia. \u003cem\u003ePragmatics Soc.\u003c/em\u003e\u003cstrong\u003e10\u003c/strong\u003e, 153\u0026ndash;176 (2019). doi: 10.1075/ps.18023.cum\u003c/li\u003e\n\u003cli\u003eBeck, A. T., Steer, R. A. \u0026amp; Brown, G. K. \u003cem\u003eBeck Depression Inventory (BDI-II)\u003c/em\u003e. vol. 10, Pearson, London, UK (1996).\u003c/li\u003e\n\u003cli\u003eRytwinski, N. K. et al. Screening for social anxiety disorder with the self-report version of the Liebowitz Social Anxiety Scale. \u003cem\u003eDepress. Anxiety\u003c/em\u003e\u003cstrong\u003e26\u003c/strong\u003e, 34\u0026ndash;38 (2009). doi: 10.1002/da.20508\u003c/li\u003e\n\u003cli\u003eRussell, D., Peplau, L. A. \u0026amp; Ferguson, M. L. Developing a measure of loneliness. \u003cem\u003eJ. Pers. Assess.\u003c/em\u003e\u003cstrong\u003e42\u003c/strong\u003e, 290\u0026ndash;294 (1978). doi: 10.1207/s15327752jpa4203_11\u003c/li\u003e\n\u003cli\u003eWeeks, D. G., Michela, J. L., Peplau, L. A. \u0026amp; Bragg, M. E. Relation between loneliness and depression: A structural equation analysis. \u003cem\u003eJ. Pers. Soc. Psychol.\u003c/em\u003e\u003cstrong\u003e39\u003c/strong\u003e, 1238\u0026ndash;1244 (1980). doi: 10.1037/0022-3514.39.6.1238\u003c/li\u003e\n\u003cli\u003eHartshorne, T. S. Psychometric properties and confirmatory factor analysis of the UCLA Loneliness Scale. \u003cem\u003eJ. Pers. Assess.\u003c/em\u003e\u003cstrong\u003e61\u003c/strong\u003e, 182\u0026ndash;195 (1993). doi: 10.1207/s15327752jpa6101_14\u003c/li\u003e\n\u003cli\u003eRussell, D. W. UCLA Loneliness Scale (Version 3): Reliability, validity, and factor structure. \u003cem\u003eJ. Pers. Assess.\u003c/em\u003e\u003cstrong\u003e66\u003c/strong\u003e, 20\u0026ndash;40 (1996). doi: 10.1207/s15327752jpa6601_2\u003c/li\u003e\n\u003cli\u003eSheehan, D. V. et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): Development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. \u003cem\u003eJ. Clin. Psychiatry\u003c/em\u003e\u003cstrong\u003e59 Suppl 20\u003c/strong\u003e, 22\u0026ndash;33; quiz 34\u0026ndash;57 (1998).\u003c/li\u003e\n\u003cli\u003eBaykara, E., Kuhn, C., Linz, N., Tr\u0026ouml;ger, J. \u0026amp; Karbach, J. Validation of a digital, tablet-based version of the Trail Making Test in the ∆elta platform. \u003cem\u003eEur. J. Neurosci.\u003c/em\u003e\u003cstrong\u003e55\u003c/strong\u003e, 461\u0026ndash;467 (2022). doi: 10.1111/ejn.15191\u003c/li\u003e\n\u003cli\u003eLinz, N. et al. Automatic detection of apathy using acoustic markers extracted from free emotional speech. In \u003cem\u003eProc. 40th Annu. Conf. Cogn. Sci. Soc.\u003c/em\u003e, 1849\u0026ndash;1855 (Cognitive Science Society, Austin, TX, 2018).\u003c/li\u003e\n\u003cli\u003eAloshban, N., Esposito, A. \u0026amp; Vinciarelli, A. What you say or how you say it? Depression detection through joint modeling of linguistic and acoustic aspects of speech. \u003cem\u003eCogn. Comput.\u003c/em\u003e\u003cstrong\u003e14\u003c/strong\u003e, 1585\u0026ndash;1598 (2022). doi: 10.1007/s12559-022-09934-4\u003c/li\u003e\n\u003cli\u003eWade-Woolley, L., Wood, C., Chan, J. \u0026amp; Weidman, S. Prosodic competence as the missing component of reading processes across languages: Theory, evidence and future research. \u003cem\u003eSci. Stud. Read.\u003c/em\u003e\u003cstrong\u003e26\u003c/strong\u003e, 165\u0026ndash;181 (2022). doi: 10.1080/10888438.2022.2028244\u003c/li\u003e\n\u003cli\u003eWu, P. et al. Automatic depression recognition by intelligent speech signal processing: A systematic survey. \u003cem\u003eCAAI Trans. Intell. Technol.\u003c/em\u003e\u003cstrong\u003e8\u003c/strong\u003e, 701\u0026ndash;711 (2023). doi: 10.1049/trit.2022.0053\u003c/li\u003e\n\u003cli\u003eGobl, C. \u0026amp; N\u0026iacute; Chasaide, A. The role of voice quality in communicating emotion, mood and attitude. \u003cem\u003eSpeech Commun.\u003c/em\u003e\u003cstrong\u003e40\u003c/strong\u003e, 189\u0026ndash;212 (2003). doi: 10.1016/S0167-6393(02)00059-3\u003c/li\u003e\n\u003cli\u003eCummins, N., Vlasenko, B., Sagha, H. \u0026amp; Schuller, B. Enhancing speech-based depression detection through gender dependent vowel-level formant features. In \u003cem\u003eArtificial Intelligence in Medicine\u003c/em\u003e, Lecture Notes in Computer Science, vol. 12501, 209\u0026ndash;214 (Springer, Cham, 2020). doi: 10.1007/978-3-030-58814-5_26\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"depression, loneliness, speech biomarker, social anxiety, speech analysis","lastPublishedDoi":"10.21203/rs.3.rs-7868485/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7868485/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eLoneliness has been demonstrated to exert a detrimental effect on mental and physical health. It may impede the formation of new social relationships by altering interactional behavior. This study provides a proof of concept that loneliness is reflected in altered speech, demonstrating that small yet significant effects can make loneliness audible. Samples of 96 healthy participants (mean age 30.85 years, 53 women) were recorded while they performed a picture description and storytelling task. Paralinguistic markers related to prosodic, formant, source, and temporal qualities of speech were extracted and correlated with loneliness, social anxiety and depression. To validate the diagnostic power, machine learning analyses were conducted for women and men separately. A model comprising all speech features from the picture description task significantly predicted loneliness. However, this model did not predict loneliness from the storytelling task. No single speech feature emerged as a strong predictor of loneliness. A combined model that included both speech features and psychiatric symptoms provided better predictions than psychiatric symptoms alone only in women. Overall, these findings suggest that speech offers a new perspective on how loneliness becomes perceptible to others and how it may disrupt social interactions, thereby fostering chronicity.\u003c/p\u003e","manuscriptTitle":"The sound of loneliness: Prediction of perceived social isolation using automatic speech analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-11 16:27:32","doi":"10.21203/rs.3.rs-7868485/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-12-31T06:01:56+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-12-15T12:53:45+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"262910393021751159413125088777895063630","date":"2025-11-17T15:53:56+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-30T12:20:19+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-10-17T17:25:45+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-16T07:12:49+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-10-16T07:08:27+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-10-15T13:04:44+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"438b5f77-f0e8-49f7-a888-8f301ca58a40","owner":[],"postedDate":"November 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":57676511,"name":"Health sciences/Health care"},{"id":57676512,"name":"Biological sciences/Psychology"},{"id":57676513,"name":"Social science/Psychology"}],"tags":[],"updatedAt":"2026-04-07T16:08:09+00:00","versionOfRecord":{"articleIdentity":"rs-7868485","link":"https://doi.org/10.1038/s41598-026-45965-5","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2026-04-04 16:00:14","publishedOnDateReadable":"April 4th, 2026"},"versionCreatedAt":"2025-11-11 16:27:32","video":"","vorDoi":"10.1038/s41598-026-45965-5","vorDoiUrl":"https://doi.org/10.1038/s41598-026-45965-5","workflowStages":[]},"version":"v1","identity":"rs-7868485","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7868485","identity":"rs-7868485","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}