"Ask" or "Inquire": operationalizing speech formality in psychosis and its risk states using etymology

Communication deficits are a hallmark of psychosis and are linked to functional impairments. This study explores the potential link between speech formality and these deficits. Previous research has shown that healthy individuals tend to use words of Latinate origin (often traced back to Old French in English) to convey formality in speech, while Germanic-origin words are associated with informality. This pattern of word choice is observed in various contexts, including everyday conversations and literary works where authors manipulate word etymology to depict character formality or intellect. This study represents the first analysis of speech etymology in a clinical cohort. The researchers developed a computational pipeline to analyze the etymology content of transcribed open-ended interviews from a large, international cohort of English-speaking individuals. This cohort comprised individuals with ROP, CHR individuals, and healthy controls (HC) matched for demographics. The researchers aimed to determine whether the proportion of Germanic and Old French words in participants' speech differed across groups and whether these differences were associated with other linguistic measures and clinical outcomes. They hypothesized that individuals with ROP would show a higher proportion of Germanic words and a lower proportion of Old French words compared to HC. They also predicted that these differences would be associated with lower lexical diversity and perplexity (word rarity). The CHR group was included to provide insights into the developmental aspects of psychosis.

The existing literature establishes a strong correlation between disordered communication and functional deficits in psychosis. Studies have demonstrated the association of formal thought disorder and speech disturbances with reduced social functioning and overall quality of life in individuals with schizophrenia. The concept of speech formality and its relationship to word etymology has been explored in linguistic studies. Research indicates a preference for Latinate words to convey formality in healthy individuals, with these words often perceived as signaling a more formal context. Analysis of literary texts reveals authors’ strategic use of Germanic versus Latinate words to portray the formality or intellectual capacity of characters. However, prior to this study, the application of etymology analysis to clinical populations experiencing psychotic symptoms remained unexplored.

This study employed a cross-sectional design involving data from two studies of language production in individuals across the schizophrenia spectrum. Participants included 92 individuals with ROP (symptom onset within 5 years), 144 CHR individuals, and 173 HC. Exclusion criteria included a risk of self-harm or harm to others, medical or neurological disorders impacting language, IQ below 70, and (for HC) a DSM Axis I diagnosis. Diagnostic assessments used the Structured Clinical Interview for DSM disorders, while CHR status was determined using the Structured Interview for Psychosis-Risk Syndromes or the Comprehensive Assessment of At-Risk Mental States, depending on the study site. Open-ended interviews, conducted using a standardized protocol, were transcribed using a HIPAA-compliant service. The transcripts were then preprocessed using NLP techniques (Stanza), including lemmatization (reducing words to their root form) and removal of special characters and punctuation. Etymology analysis was performed using two resources: Etymonline.com and a database derived from Wiktionary.com. The proportion of Germanic and Old French words (restricted to nouns, adjectives, verbs, and adverbs) was calculated for each transcript. Lexical diversity was measured using Honoré's Statistic, while perplexity (word rarity) was calculated using word frequencies from the Google N-grams database. Statistical analyses included ANOVAs, t-tests, correlation analyses, and multiple linear regressions to examine group differences in etymology proportions, lexical diversity, and perplexity, adjusting for covariates such as sex, age, recruitment site, education, race, and (in a subset) IQ. Associations between etymology, clinical measures (symptom severity and functioning), and other linguistic variables were also investigated.

The researchers found a strong correlation between etymology proportions derived from Etymonline.com and Wiktionary.com, opting to use Wiktionary data for further analyses due to its multilingual capabilities. After adjusting for covariates, healthy controls (HC) demonstrated significantly lower proportions of Germanic-origin words and higher proportions of Old French-origin words compared to individuals with recent-onset psychosis (ROP). This pattern also held true for individuals at clinically high risk (CHR) for psychosis, who showed similar etymology patterns to the ROP group. Consistent with hypotheses, HC showed greater lexical diversity than both clinical groups. However, perplexity did not differ significantly between groups. These findings remained consistent in the subgroup with available IQ data. Analyses within recruitment sites showed largely similar results, though differences in Germanic word use were only trend-level significant in New York. Significant differences in lexical diversity were only identified in Melbourne, with the ROP group exhibiting lower diversity. All lexical features (Germanic word use, Old French word use, lexical diversity, and perplexity) were significantly intercorrelated. Multivariate regression models indicated that clinical cohort membership, lexical diversity, and perplexity significantly contributed to the variance in both Germanic and Old French word use, while other sociodemographic factors had less impact. Importantly, Old French word use was negatively associated with negative symptoms only in the ROP group, but this association did not survive Bonferroni correction. Finally, and importantly, increased Old French word use (and decreased Germanic word use) significantly correlated with better role functioning (but not social functioning) in the combined clinical cohort, an association robust to Bonferroni correction.

The study's findings support the hypothesis that individuals with ROP use proportionately more Germanic-origin words and fewer Old French-origin words than healthy controls, indicating less formal speech. This pattern extends to individuals at CHR for psychosis. While lexical diversity and perplexity were correlated with etymology, they did not fully account for group differences in speech formality. The unique association of speech formality (indexed by etymology) with role functioning, but not social functioning, suggests that this linguistic feature may be specifically relevant to professional or academic contexts. These differences in speech formality may reflect variations in cognitive load management (as captured by lexical diversity and perplexity), with individuals with psychotic symptoms potentially opting for simpler, more readily accessible words. The study highlights the potential of analyzing speech etymology as a method for operationalizing speech formality in psychosis.

This study pioneers the investigation of speech etymology content in a clinical cohort, revealing differences in Germanic and Old French word use between individuals with psychotic symptoms and healthy controls. These differences are partially explained by variations in lexical diversity and speech perplexity. However, speech etymology provides a unique measure of formality linked to impaired role functioning in psychosis. Future research should explore the relationship between speech formality, cognition, and functional outcomes more comprehensively, including the use of cognitive batteries and detailed antipsychotic dosage data. Further research could also investigate the generalizability of these findings across other populations and languages.

Several limitations exist in this study, particularly concerning data availability. IQ data were available only for a subset of participants, and concurrent cognitive data (from the MATRICS Consensus Cognitive Battery) were also limited due to the COVID-19 pandemic. Detailed antipsychotic dosage information was unavailable for participants with ROP. Data on native language, language spoken at home, and migration history, which could affect vocabulary and the association of English with formality, were also lacking. Future studies should address these limitations by including more comprehensive cognitive assessments, antipsychotic dosage information, and detailed language acquisition data.