Tracking the acceptance of neologisms in German: Psycholinguistic factors and their correspondence with corpus-linguistic findings

The study examines whether neologisms borrowed from English are accepted more slowly or less readily by German speakers than neologisms formed from native German material (primarily compounds). The context spans lexicology and sociolinguistics of borrowing, where integration involves morpho-syntactic and semantic adaptation and a sociological process of acceptance. Prior corpus work suggested that borrowed neologisms are initially marked more often (e.g., quotation marks or hedges) but become less marked over time, implying slower acceptance compared to native formations. The current research tests this assumption psycholinguistically, using mouse-tracking to infer decision uncertainty when participants judge whether a word "is being used" in German. The purpose is to provide converging evidence between corpus indicators (frequency and linguistic flagging) and experimental measures of acceptance and processing, including potential age-related differences. Hypotheses: (H1) Borrowed neologisms will have lower acceptance rates and longer reaction times than native ones; (H2) accepted borrowed neologisms will show more mouse-trajectory uncertainty than accepted native ones; (H3) these effects should be attenuated or absent for neologisms from the 1990s due to greater entrenchment.

The paper reviews perspectives on lexical borrowing. Prescriptive traditions often resist loanwords and advocate purification of the lexicon, while descriptive approaches view borrowings as legitimate enrichments potentially used for prestige or specific functions. Integration criteria include linguistic (morpho-syntactic, semantic), sociological factors, and dictionary inclusion as a norm indicator. Studies show younger speakers may use borrowings more frequently than older ones. Corpus-based work (Klosa-Kückelhaus & Wolfer, 2020) on 239 German neologisms from the 1990s found no clear frequency difference in acceptance between borrowed and native items, but borrowed neologisms were more frequently flagged early on, with flagging decreasing over time. Other research highlights accessibility issues with borrowings leading to flagging, dispersion/frequency considerations as predictors of uptake, and the importance of sociopragmatic parameters and attitudes in the selection of Anglicisms versus native alternatives. The notion of neological intuition shows that recognition of novelty can vary with neologism type and formation regularity.

Participants: Data from 66 native German speakers (after exclusions) collected July–December 2022. Age range 15–85 years (mean 36.7; median 25.5). Grouped by birth year: 44 born in 1980 or later (millennials) and 22 born before 1980. Gender not recorded. All testing used the same Windows 10 laptop with external mouse and headphones.

Stimuli and design: Noun neologisms from the IDS Neologismenwörterbuch for the 1990s and 2010s, either borrowed from English or formed from German material. All German-formed items were compounds; most English-origin items were compounds (in German orthography written concatenated; some with hyphen) and some derivatives. All English-origin test words are attested in English (no pseudo-Anglicisms). Corpus frequency in DeReKo was matched and also entered as a covariate. Each participant judged 24 items across 6 conditions in a 2×3 design: Origin (English vs German) × Time/Status (1990s, 2010s, Pseudo). Pseudo-neologisms conformed to German formation patterns but were unattested, to permit legitimate rejections. All items were presented to all participants in randomized order with at most two consecutive trials from the same condition. Visual presentation was accompanied by an audio recording (same native speaker) adapted to German phonology.

Procedure: MouseTracker software recorded trajectories. Trial structure: click START at bottom center; then word appears centrally and audio plays; two response buttons JA/NEIN appear in upper corners (left–right positions counterbalanced across participants). Participants responded to the prompt: "Ist das ein Wort, das verwendet wird?" (“Is this a word that is being used?”). Four practice trials preceded experimental trials; a brief break occurred midway.

Data processing: Trials with initiation time >1 s excluded (4.55%). Among remaining, trials with log RT outside mean ±3 SDs excluded (1.26%). Final dataset: 8,958 trials. Analyses focused on non-pseudo items (pseudo rejection rate 76.5% overall; English 73.2%, German 79.7%); non-pseudo rejections were 27.2%. Non-pseudo dataset: 5,988 trials. Mouse trajectories were remapped, aligned, and time/space-normalized (101 time steps) using mousetrap. Dependent variables: acceptance (yes/no), log RTs (analyzed for accepted trials), x-axis flips (direction changes), maximum absolute deviation (MAD) from ideal path, average deviation (AD); AUC reported similarly in supplement. Outliers per trajectory measure were removed at mean ±3 SD after prior exclusions.

Modeling: Mixed-effects models (lme4). Starting models included random intercepts for participants and items, by-participant random slopes for trial position; fixed effects included Origin, Time (1990s vs 2010s), Age group (≥1980 vs <1980), all interactions, plus covariates word length and log corpus frequency (scaled), and trial position (scaled). Models were reduced via likelihood ratio tests to include only contributors to fit. Reported are fixed estimates with 95% CIs, random effect variances, and marginal/conditional R². Reaction-time and trajectory analyses were restricted to accepted trials.

• Acceptance: Contrary to H1, German neologisms had a lower estimated probability of acceptance than English-origin neologisms. A significant three-way interaction Time × Origin × Age group was present. Younger participants (born ≥1980) showed a pronounced disadvantage for German neologisms; for older participants, English neologisms from the 2010s were slightly less acceptable than those from the 1990s. Higher corpus frequency increased acceptance.
• Reaction times (accepted trials): Responses were slower for German neologisms compared to English, particularly among younger participants. Younger participants were overall faster (Age group ≥1980 negative estimate). Longer words increased RT; higher corpus frequency and later trial position decreased RT. A Time × Age interaction suggested minor differences across decades, but pairwise contrasts were not interpretable.
• Mouse-trajectory uncertainty (accepted trials): For x-axis flips, MAD, and AD, younger participants showed greater uncertainty (more flips; larger curvature/deviation) when accepting German neologisms than English ones. No reliable Origin effects were found in these measures for the older group. Higher corpus frequency lowered flips; trial position modestly affected MAD/AD patterns; longer words slightly reduced AD. Overall, uncertainty patterns invert H2: English neologisms were accepted with less uncertainty by millennials.
• Pseudo items check: Pseudo-neologisms were rejected in 76.5% of trials (English 73.2%, German 79.7%), confirming task sensitivity. Overall, results run counter to corpus-derived expectations: English-origin neologisms were accepted more often, more quickly, and with less uncertainty than German-formed neologisms, but primarily among millennials.

Findings challenge assumptions derived from corpus flagging behavior. The expected disadvantage for borrowed neologisms (lower acceptance, slower, greater uncertainty) was not observed; instead, English-origin items enjoyed an advantage among younger speakers, with negligible Origin effects among older participants. This mismatch suggests that linguistic flagging in written corpora may not straightforwardly index community-wide acceptance; flagging reflects individual authors’ editorial choices and accessibility-marking strategies and may not map onto speakers’ real-time judgments. Potential explanations include: (1) Decision strategies differ by Origin: participants reported being more lenient toward English borrowings, accepting them when they felt familiar, whereas German neologisms prompted effortful semantic retrieval, increasing decision time and trajectory complexity. (2) Corpus frequency progression (temporal trends) rather than overall frequency may better predict acceptance; previous corpus analyses showed English borrowings maintaining higher levels while German formations declined over time. However, this does not fully explain the age-specific advantage. (3) Exposure and subjective frequency: millennials may encounter English-origin neologisms more broadly and frequently in daily life, leading to faster, more confident acceptance. Measures of dispersion across subcorpora (DPnorm) did not differ between groups, and domain labels did not account for effects. The decade of neologism introduction affected acceptance for older participants but had limited impact otherwise. Together, results argue for a nuanced interpretation of corpus indicators and underscore the role of age-related experience and attitudes in processing neologisms.

The study shows that, especially among millennials, English-origin neologisms in German are accepted more often, processed faster, and with less uncertainty than German-formed neologisms. This contradicts corpus-based expectations derived from linguistic flagging and suggests that flagging is an imperfect proxy for acceptance. Future work should incorporate measures of language experience and attitudes, evaluate frequency trajectories and dispersion more directly, and assess English proficiency as a potential moderator. Extending beyond isolated word judgments to contextualized usage and alternative paradigms (e.g., sentence embedding, prevalence surveys) could clarify mechanisms driving acceptance and integration of neologisms.

• Participant sample skewed toward younger adults; relatively small older group may limit power for age comparisons.
• English proficiency was not measured; proficiency could modulate acceptance of English-origin neologisms.
• Only nouns were tested, mostly compounds; results may not generalize to other parts of speech or formation types.
• Stimuli differed across conditions (no item-level matching across all cells), which may introduce item-specific effects despite random effects and covariate controls.
• Tasks used isolated words without context; processing in context may differ.
• Self-reported debrief insights about decision strategies were anecdotal and not systematically collected.
• Corpus covariates included overall frequency, not full temporal frequency trajectories at the item level within models.