It's a small world of signs: Semantic network structure in American Sign Language based on free associations
Abstract
Semantic network analyses have shown that words in spoken languages are organized into efficient "small-world" structures that support rapid lexical access and retrieval. However, little is known about whether similar organizational principles generalize to sign languages. We present the first large-scale analysis of semantic network structure in American Sign Language (ASL) using a free semantic association task. Deaf ASL signers (N = 45) produced three associations to 2,714 cue signs drawn from ASL-LEX, resulting in almost 100,000 cue–response pairs. We constructed (a) directed associative networks based on cue-response relationships and (b) undirected conceptual similarity networks based on cue–to-cue cosine similarity and compared ASL with English using matched subsets of the Small World of Words norms. We estimated network statistics and evaluated network topology using bootstrapping and degree-preserving random graphs. Both languages exhibited small-world organization indicating similar pressures on efficient semantic organization across language modalities. However, the languages differed in topology. Across network types, ASL exhibited longer average path lengths, and in conceptual similarity networks, lower clustering, suggesting reduced integration. In associative networks, ASL showed lower modularity, while in the conceptual similarity networks, ASL showed higher modularity than English, indicating a more compartmentalized semantic space. Further, more frequent signs occupied denser and more diverse semantic neighborhoods, whereas iconic signs exhibited fewer semantic connections and lower response diversity. These findings suggest that semantic memory is shaped by modality-independent organizational principles while also reflecting language-specific network topologies and structural pressures. The dataset and tools provide a basis for future cross-linguistic investigations.
Related
- Semantic Cognition for and from Language Models
- Assessing the Impact of Language on Visual Threat Detection: cTBS over the anterior temporal lobe impairs semantic processing but not visual threat detection
- Analysis and improvement of surface electromyography-based American Sign Language hand gesture recognition
- Lifespan accumulated differences in semantic memory structure contribute to increases in episodic false memories
- ACMNet: An attention-convolution mixed neural network for semantic segmentation of high-resolution remote sensing images
BibTeX
@misc{sehyr2026small,
title = {{It's a small world of signs: Semantic network structure in American Sign Language based on free associations}},
author = {Zed Sehyr and Karen Emmorey and Elana Pontecorvo and Naomi Caselli and Erin Michelle Buchanan},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/uejts_v1},
url = {https://osf.io/preprints/psyarxiv/uejts_v1/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - It's a small world of signs: Semantic network structure in American Sign Language based on free associations
AU - Zed Sehyr
AU - Karen Emmorey
AU - Elana Pontecorvo
AU - Naomi Caselli
AU - Erin Michelle Buchanan
PY - 2026
DA - 2026/09/26
PB - PsyArXiv
DO - 10.31234/osf.io/uejts_v1
UR - https://osf.io/preprints/psyarxiv/uejts_v1/
AB - Semantic network analyses have shown that words in spoken languages are organized into efficient "small-world" structures that support rapid lexical access and retrieval. However, little is known about whether similar organizational principles generalize to sign languages. We present the first large-scale analysis of semantic network structure in American Sign Language (ASL) using a free semantic association task. Deaf ASL signers (N = 45) produced three associations to 2,714 cue signs drawn from ASL-LEX, resulting in almost 100,000 cue–response pairs. We constructed (a) directed associative networks based on cue-response relationships and (b) undirected conceptual similarity networks based on cue–to-cue cosine similarity and compared ASL with English using matched subsets of the Small World of Words norms. We estimated network statistics and evaluated network topology using bootstrapping and degree-preserving random graphs. Both languages exhibited small-world organization indicating similar pressures on efficient semantic organization across language modalities. However, the languages differed in topology. Across network types, ASL exhibited longer average path lengths, and in conceptual similarity networks, lower clustering, suggesting reduced integration. In associative networks, ASL showed lower modularity, while in the conceptual similarity networks, ASL showed higher modularity than English, indicating a more compartmentalized semantic space. Further, more frequent signs occupied denser and more diverse semantic neighborhoods, whereas iconic signs exhibited fewer semantic connections and lower response diversity. These findings suggest that semantic memory is shaped by modality-independent organizational principles while also reflecting language-specific network topologies and structural pressures. The dataset and tools provide a basis for future cross-linguistic investigations.
ER -