Memories that remain: Image memorability transfers to scene descriptions

Abstract

Some images remain in memory while others fade quickly. Memorability reflects an observer-independent, intrinsic property of an image, which can be predicted by computational models. While semantic content has been proposed as the critical determinant of memorability, it is challenging to disentangle image content from visual features on display. Here, we find that the memorability of scene images transfers to short textual descriptions of those scenes, whether generated by humans or multimodal large language models. Furthermore, new images produced with generative AI from descriptions alone reproduce the memorability of original images. Together, these findings show that semantic content, including interrelations captured in scene descriptions, accounts for a meaningful portion of the variation in scene memorability, and that language serves as an efficient compression of memory-relevant visual content.

The Impact of Noise on Memorability in Minds and Machines
H. Ding, M. Wiener
bioRxiv19 Oct 2026PreprintCC BYFull textNeurosciencebioRxiv
Degree-based and functional descriptions of pediatric hearing loss: A retrospective chart review
Caitlin Sapp, Allison Duplain, Ryan W. McCreery +2
International journal of audiology20 Sep 2026Peer-reviewedLicence not statedFull textNeurosciencePMC
Review of Dragon Millipedes (Diplopoda, Polydesmida, Paradoxosomatidae) in the Fauna of Vietnam, with Descriptions of Three New Species
Anh Duc Nguyen, Manh-Ha Nguyen, Thu-Anh T Nguyen +1
PubMed Central10 Sep 2026Peer-reviewedLicence not stated5 citedEcologyOpenAlex
The extensive Indo-Pacific deep-water radiation of Vermeijius Kantor, Fedosov, Snyder & Bouchet, 2018 (Buccinoidea: Fasciolariidae), revealed by molecular data, with descriptions of new species
Roland Hadorn, Yuri I Kantor, Philippe Bouchet +1
European Journal of Taxonomy18 Sep 2026Peer-reviewedCC BY 4.0Full textEcologyHAL
A New Genus of Oak Gall Wasp, Striatoandricus Pujade-Villar (Hymenoptera: Cynipidae: Cynipini) from America with Descriptions of Two New Mexican Species
Víctor Cuesta-Porta, Miquel A. Arnedo, David Cibrián-Tovar +7
PubMed Central10 Sep 2026Peer-reviewedLicence not stated10 citedEvolutionary biologyOpenAlex

Cite

BibTeX RIS

BibTeX

@misc{kollenda2026memories,
  title = {{Memories that remain: Image memorability transfers to scene descriptions}},
  author = {Diana Kollenda and Yee Lee Shing and Katharina Dobs and Benjamin de Haas},
  howpublished = {PsyArXiv},
  year = {2026},
  month = aug,
  doi = {10.31234/osf.io/mg7zk_v2},
  url = {https://osf.io/preprints/psyarxiv/mg7zk_v2/},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Memories that remain: Image memorability transfers to scene descriptions
AU  - Diana Kollenda
AU  - Yee Lee Shing
AU  - Katharina Dobs
AU  - Benjamin de Haas
PY  - 2026
DA  - 2026/08/31
PB  - PsyArXiv
DO  - 10.31234/osf.io/mg7zk_v2
UR  - https://osf.io/preprints/psyarxiv/mg7zk_v2/
AB  - Some images remain in memory while others fade quickly. Memorability reflects an observer-independent, intrinsic property of an image, which can be predicted by computational models. While semantic content has been proposed as the critical determinant of memorability, it is challenging to disentangle image content from visual features on display. Here, we find that the memorability of scene images transfers to short textual descriptions of those scenes, whether generated by humans or multimodal large language models. Furthermore, new images produced with generative AI from descriptions alone reproduce the memorability of original images. Together, these findings show that semantic content, including interrelations captured in scene descriptions, accounts for a meaningful portion of the variation in scene memorability, and that language serves as an efficient compression of memory-relevant visual content.
ER  -

Abstract

Related

Cite