Across the deep seafloor, the formation of large holes and cave-like structures in fine-grained sediments has been attributed to powerful geological forces. Observations from the continental rise off Lofoten, Northern Norway, at depths of around 2200 m have documented extensive systems of such caves (Bellec et al., 2010; Bøe et al., 2012). The prevailing scientific explanation posits a geological origin, suggesting they are “venting-tunnels” or “pockmark caverns” created by the catastrophic escape of fluids and gas, such as methane, from the sediment (Hovland and Judd, 1988; Bøe et al., 2012). These events are thought to be triggered by pressure changes that destabilize gas hydrates, leading to a violent fluidization of sediments and the creation of complex subterranean conduit systems (Hovland et al., 2002).

This study presents new observations that challenge this purely geological explanation. During remotely operated vehicle (ROV) surveys in the Central Mohns Ridge area, we discovered similar cave structures that were sometimes occupied by the glacial eelpout (Lycodes frigidus Collett, 1879, family Zoarcidae). This benthic deep-sea fish is widely distributed in the Arctic Ocean and common in the Norwegian Sea, yet little is known about its behavior and life cycle (Møller and Jørgensen, 2000; Mecklenburg et al., 2018). It typically occurs at depths between 475–3580 m in water temperatures of -0.6 to 1.6°C (Hildebrandt et al., 2011; Mecklenburg et al., 2018), and feeds on benthic fauna (Bjelland et al., 2000) while being preyed upon by the Arctic skate [Amblyraja hyperborea (Collett, 1879)] (Byrkjedal et al., 2015). Spawning is believed to take place in autumn or winter (Prouse and McAllister, 1986; Hildebrandt et al., 2011).

Fish can significantly alter the seabed topography through impressive digging behavior [e.g. pufferfish (Kawase et al., 2013), tilefish (Twichell et al., 1985), and icefish (Purser et al., 2022)]. The use of burrows and other structures for reproduction and shelter is well documented for the family Zoarcidae (Kendall et al., 1983; Ferry-Graham et al., 2007), and nesting behavior has recently been suggested for another Lycodes species (Kennedy et al., 2024). Based on these observations, we present an alternative hypothesis that L. frigidus may contribute to the creation or modification of deep-sea caves, along with the first documented case of egg-guarding by this species.

During the DeepInsight cruise with RV Kronprins Haakon in March 2025, surveys were conducted in the Greenland and Norwegian Sea as part of the EMINENT project (eminent-project.com). The objectives of the cruise were to map potential seafloor massive sulfide (SMS) resources and associated biology through physical sampling and ROV video surveys. The ROV ÆGIR6000 was deployed, equipped with a 4K video camera, HD cameras, LED lighting, two manipulators, and parallel laser pointers (10 cm separation). The laser pointers, as well as photogrammetry models based on the ROV video material, were used for estimating sizes of objects of interest. The photogrammetry results were compared with the laser distances, and the maximum deviation was found to be ± 5%. Depending on the model mesh resolution, which is variable throughout the model, we can determine distances down to a few centimeters. The position of the ROV was determined with an accuracy of 5–20 m using an integrated navigation system (HiPAP, DVL, and INS) (Pedersen et al., 2025). The observations presented here are from dives 22 and 23 in the Grøntua and DeepInsight Hill areas. Additionally, similar structures have been observed in ROV video material from previous cruises, conducted by the Norwegian Offshore Directorate (NOD) in 2024 and the EMINENT project in 2023.

At approximately 1600 m depth (72.4924°N, 1.5054°E), nine cave-like structures were observed in fine-grained sediments ( Figure 1 ). The cave entrance widths ranged from 26 cm to 170 cm, narrowing inwards, and the caves extended down to depths ranging from 50 to 100 cm ( Figure 1F ). Importantly, the back-walls of the caves, where visible, showed no evidence of rupture or fracturing, as laminae in the sediment could be traced continuously in 3D ( Figures 1A, F ). There were no signs of mineralizations, nor was there any evidence of bacterial mats associated with the caves. A single L. frigidus was observed in caves 2 ( Figures 1A, B ) and 6 ( Figures 1D, E ), with one guarding a pile of white eggs that were partially hidden behind a rock ( Figure 1B ). The fish total length (TL) was estimated to 38 cm for the one guarding the eggs and 36 cm for the other fish. Numerous small holes (approximately 1–2 cm diameter) were visible on cave walls and the surrounding seafloor, with some having small anemones protruding from them. Some caves had anemones on the seafloor extending almost as far as could be seen into the cave. Others had dislodged single sponges or piles of different sponges at the cave entrance and inside the cave ( Figures 1A, C, D ). The sponges represented sponges that were observed nearby in the area, including glass sponges in the family Rossellidae, demosponges in the order Tetractinellida, and carnivorous sponges (family Cladorhizidae), all at various stages of decomposition. Associated with these piles were also living shrimps, amphipods, anemones, and sea stars.

The observation areas from Dive 22 and 23 are part of a several-kilometer-wide area where these caves are present, located approximately 20 km from the Mid-Ocean Ridge (MOR) on oceanic crust aged ~3.5 Ma (see Supplementary Figure S1 for an overview map). A cave-like structure from Dive 22, observed at approximately 1200 m depth (72.5159°N, 1.4526°E), was located about 3.2 km north-west of the main cluster of caves from Dive 23 ( Figure 2A ). The cave entrance was estimated to be 60 cm wide. An ROV manipulator with a camera was inserted into the cave to inspect the interior ( Figure 2B ). An individual L. frigidus (TL 50 cm) was observed on the sediment about 10 m from the cave ( Figure 2C ). An Amblyraja hyperborea (TL 85 cm) was observed about 65 m away ( Figure 2D ).

Our observations provide compelling evidence for a biological role in the origin of deep-sea sedimentary caves, challenging the exclusive geological model. The catastrophic fluid escape hypothesis is typically associated with specific geomorphological evidence, such as fields of pockmarks, evidence of sediment fluidization, or distinct vertical conduits leading to horizontal tunnels (Hovland and Judd, 1988). The caves we observed did not appear to be part of a larger pockmark field and lacked the chaotic or eruptive features expected from a violent fluid blowout. Instead, they appeared as stable, established structures, suggesting a more persistent formation and maintenance process. There is no morphological evidence that the caves were formed by fluid escape. Most notably, the caves are deep and have ‘roofs’, contrasting with fluid-escape driven pockmarks that are shallow (large diameter to depth ratio), and venting systems that typically cause cones protruding from the seafloor.

In addition to the morphological criteria, there are no obvious reasons why the holes could be made by fluid escape. The sedimentary layer is thin, and crystalline outcrop is common in the area, so compaction-driven porewater extrusion can be excluded. There is no biogenic source for organic carbon, which is the common origin for e.g. methane expulsion in hydrocarbon provinces at shallow to intermediate water depth (Hovland and Judd, 1988; Sen et al., 2019; Gupta et al., 2022). Nor is there any evidence for the bacterial mats typical for cold seeps and hot vents (Sen et al., 2019; Früh-Green et al., 2022). Furthermore, the pressure-temperature conditions lie far within the hydrate stability field (Schmidt et al., 2022) for hydrocarbon vapor, and CO₂ will be supercritical (Engineering ToolBox, 2001) and thus have a density similar to seawater. Finally, the hydrothermal venting of H₂O, H₂, H₂S, CO₂ and CH₄ typical for the mid-oceanic ridge system, and its associated mineralization and vent fauna (e.g., Pedersen et al., 2010; Früh-Green et al., 2022) bears no resemblance to these off-axis caves.

We propose that Lycodes frigidus actively create or modify these caves. This argument is supported by several lines of evidence. Firstly, the discovery of an individual guarding eggs is a critical behavioral observation. Parental care is known in the Zoarcidae family, and using a secure burrow is a key reproductive strategy for many fish (Blumer, 1982). This is the first direct evidence of such nesting behavior in L. frigidus, though it has been inferred for another Lycodes species (Kennedy et al., 2024) and observed in other deep-sea species in the family Zoarcidae from the Pacific. Even midwater zoarcids are believed to descend to the seafloor to burrow their eggs in sediments (Ferry-Graham et al., 2007).

Secondly, the caves likely function as a refuge from predation. For example, the Arctic skate (Amblyraja hyperborea) is a known predator of L. frigidus, being able to consume L. frigidus up to 80% of their own TL (Byrkjedal et al., 2015). The two species have overlapping habitats (Hildebrandt et al., 2011; Mecklenburg et al., 2018) and the observation of A. hyperborea on Dive 22 confirms that they both occur in the studied area. Given the skate’s body dimensions with a broad, thorn-covered disc [~0.76 x TL (Kulka et al., 2024)] and TL up to 112 cm (Last et al., 2016), it would be unable to access the narrow cave entrances, providing a secure shelter for the eelpouts and their eggs.

Thirdly, large-scale bio-excavation by fish, while remarkable, is not unprecedented. Male pufferfish in the genus Torquigener Whitley, 1930, for instance, use their body and fins to create intricate geometrical structures on the seafloor. These structures serve both as courtship displays to attract females and as nesting sites where the males care for the eggs post-spawning. Notably, these constructions require several days to complete and are disproportionately large relative to the fish themselves (approximately 2 meters in diameter, compared to a TL of ~120 mm) (Kawase et al., 2013; Bond et al., 2020). The tilefish Lopholatilus chamaeleonticeps Goode & Bean, 1879 is known for constructing large funnel-shaped burrows on the continental shelf by moving sand and clay, further facilitating burrowing by decapods in the upper part of the burrows (Twichell et al., 1985). These structures can alter the local seafloor, much like the caves we observed. Given these known analogues, it is plausible that L. frigidus is capable of similar excavation. The numerous small surrounding holes, resembling those made by the amphipod Neohela S. I. Smith, 1881 [e.g., Neohela monstrosa (Boeck, 1861) (Buhl-Mortensen et al., 2016)], likely confirm that the local sediment is amenable to bio-excavation. Caves along the continental slope of Lofoten with similar morphology, and also having decaying biological debris at their entrances, have previously been interpreted to result from fluid escape (Bellec et al., 2010) but may also have the biological origin proposed here. To provide a broader context on the distribution of these structures, we have compiled the known occurrences in a supplementary maps ( Supplementary Figure S1 ). These maps show our primary study sites, additional caves observed along the ROV tracks of Dive 22 and 23, the area previously described by Bøe et al. (2012), and a newly identified location approximately 50 km from the primary sites (termed Cave C in the Supplementary Material ), located about 42 km off-axis on oceanic crust aged ~7.5 Ma. While a systematic survey has yet to be conducted and our observations only relate to selected, analyzed ROV video material, this compilation demonstrates that such caves are not isolated features but are found in multiple locations across the region. This wider distribution supports a common biological origin, likely linked to the frequent occurrence of L. frigidus, rather than site-specific geological events.

The cave observations documented here not only shed light on how they are made and the parental behavior of L. frigidus, but also present other interesting observations such as the piles of dislodged sponges in front of some of the caves. It is surprising to see so many dislodged sponges, and this raises questions regarding the time period between dislodging and our observations. Another question is whether these sponges are found because local currents move them and they get “trapped” by some caves - no obvious pattern could be inferred. Another possibility could be that these sponges are harvested by L. frigidus and placed at the cave entrances to hide the cave or for courtship reasons. We also observe that the cave shown in Figure 2B appears to have collected some fluffy material at the cave bottom, including a feather, while the caves that are occupied by L. frigidus do not show this fluffy material accumulation, possibly hinting that L. frigidus actively cleans and maintains these caves.

To our knowledge, what we present is the first observation of parental care in L. frigidus through the use of caves and guarding of eggs. Guarding of eggs by the female or both parents is a known parental care behavior for the family Zoarcidae (Blumer, 1982). Our observation of eggs in March, combined with literature suggesting autumn or winter spawning (Prouse and McAllister, 1986; Hildebrandt et al., 2011) points to a prolonged incubation period of several months. This is consistent with the reproductive strategies of other deep-sea Zoarcids, which have large eggs and slow development rates in cold water (Ferry-Graham et al., 2007). In conclusion, while a geological process may initiate some seafloor depressions, the evidence suggests that the structure, maintenance, and ecological function of these caves could be biologically driven by L. frigidus.