Galactic Constellations in Cosmology

Updated 5 July 2026

Galactic constellations are visually distinctive, subjective patterns in the large-scale galaxy distribution that serve as analogs to traditional stellar constellations.
They are identified through human inspection of high-completeness, thin declination slices in DESI DR1, rather than by formal clustering algorithms.
The approach uses visual encoding of galaxy brightness and comoving coordinates to reveal coherent structures like Pisces Grandis and the DESI Stick Woman, offering novel cosmological insights.

Galactic constellations are human-recognizable shapes traced by galaxies in large cosmological surveys, introduced in a DESI DR1 study as “charming shapes in large cosmological surveys” rather than as formally defined physical structures (Lamman, 31 Mar 2026). In this usage, the term denotes subjective, aesthetically chosen patterns that stand out in slices of the galaxy distribution in comoving space. The concept is deliberately analogous to traditional stellar constellations on the celestial sphere, but it is applied to the large-scale structure of the nearby Universe. The DESI work treats these shapes both as a mode of visual engagement with the cosmic web and as an unconventional, qualitative probe of cosmological homogeneity (Lamman, 31 Mar 2026).

1. Terminological scope and astronomical context

In astronomy, “constellation” has more than one established usage. The 88 IAU constellations segment the sky into exact regions separated by official boundaries, whereas many practical and cultural discussions focus on visually salient star groupings or asterisms. The IAU boundary system is computationally awkward because the boundaries follow a complicated network of lines of constant right ascension and declination at epoch B1850.0, drawn in a “winding” way so that older historical assignments would remain consistent (Glaschke, 2010). This exact, region-based meaning differs fundamentally from the DESI notion of galactic constellations.

Galactic constellations borrow the visual and mnemonic aspect of traditional constellations rather than the formal boundary-based one. They are not sky regions, and they are not catalogued dynamical entities such as halos, filaments, or spiral arms. The DESI DR1 study explicitly emphasizes that they are subjective patterns found in a dense galaxy distribution and that this subjectivity is part of the point: they are meant to help observers intuit the scale and texture of the cosmic web (Lamman, 31 Mar 2026).

This distinction is important because it prevents a category error. Traditional constellations, in the IAU sense, support exact positional classification; galactic constellations support human pattern recognition in cosmological data. A plausible implication is that the DESI terminology is best understood as an observational metaphor grounded in visual cognition rather than as a new taxonomic class of large-scale-structure object.

2. Perceptual grouping and the analogy to stellar asterisms

Cross-cultural work on stellar asterisms provides a rigorous framework for understanding why recognizable patterns recur in point distributions. A study of 27 constellation systems—22 from the Stellarium software package and 5 from ethnographic literature—showed that 28% of 605 asterisms recur across cultures when recurrence is defined by a human score threshold of at least 0.2 (Kemp et al., 2020). The most frequent recurring asterisms include Orion’s Belt, the Pleiades, the Hyades, the Big Dipper, the Southern Cross, and Cassiopeia, as well as Corona Borealis, Delphinus, the head of Aries, the Southern Pointers, the head/stinger/tail of Scorpius, the sickle in Leo, the shaft of Aquila, Corvus, Lyra, the Northern Cross, the Square of Pegasus, Corona Australis, and the head of Draco (Kemp et al., 2020).

That study explains recurrence through a Graph Clustering model of perceptual grouping based on proximity and brightness. The model uses the union of three Delaunay triangulations over stars brighter than 3.5, 4.0, and 4.5 magnitudes; combines proximity and brightness with $\rho = 3$ ; rescales salience within a $60^\circ$ neighborhood; and identifies asterisms as connected components after retaining the $n$ strongest edges. Under a model-score threshold of at least 0.2, it captures 98% of the recurring asterisms and 80% of the entire dataset (Kemp et al., 2020).

A plausible implication is that galactic constellations exploit related perceptual mechanisms, even though the underlying points are galaxies in comoving slices rather than stars on the celestial sphere. The DESI study does not formalize that connection as a perceptual model, but its emphasis on visually coherent, memorable shapes is consistent with the broader result that human observers do not select point patterns arbitrarily.

3. Identification in DESI DR1

The first galactic constellations in the DESI study were identified by visually inspecting a dense subset of DESI DR1. The survey was cut to the region with the highest completeness, between $125 < \alpha < 250$ degrees in right ascension and $-7 < \delta < 7$ degrees in declination, and limited to $z < 0.8$ , a sample dominated by Bright Galaxy Survey galaxies with some Luminous Red Galaxies (Lamman, 31 Mar 2026). To render structure more legible, the authors projected the data into thin declination slices, usually 2 degrees thick.

The visualization strategy was part of the method. The galaxies were plotted with semi-transparent symbols, and brightness was scaled with declination so that foreground and background structure could be distinguished. Positions were converted into comoving coordinates using redshifts and Planck 2018 cosmological parameters, so the size of a galactic constellation corresponds to a physical distance rather than a purely angular extent (Lamman, 31 Mar 2026).

These choices define the operational meaning of the term. Galactic constellations are not extracted by a clustering algorithm and are not inferred from a dynamical model; they are found by human inspection of carefully selected, high-completeness slices in comoving space. This makes the method intrinsically observer-centered. It also makes the resulting objects useful as a visual probe of what scales still support coherent pattern recognition in the cosmic web.

4. Named examples and participatory discovery

The DESI DR1 study highlights three named examples: Pisces Grandis, The DESI Stick Woman, and W (Lamman, 31 Mar 2026). Pisces Grandis is described as “the largest known fish in the Universe,” about $270\,h^{-1}\mathrm{Mpc}$ long, or roughly 900 million light years. The DESI Stick Woman is presented as a “potential friend” of the CfA Stick Man from the 1985 CfA Redshift Survey. W is a letter-like pattern whose resemblance to Cassiopeia is explicitly noted (Lamman, 31 Mar 2026).

Constellation	Description in the DESI study	Reported scale or comparison
Pisces Grandis	“the largest known fish in the Universe”	about $270\,h^{-1}\mathrm{Mpc}$ long; roughly 900 million light years
The DESI Stick Woman	whimsical counterpart to the CfA Stick Man	presented as a “potential friend”
W	striking letter-like pattern	remarked to resemble Cassiopeia

The paper also created a public website, cmlamman.github.io/galactic-constellations, to systematize exploration and sharing. The site serves random DESI galaxy slices from the same DR1 region, each 2 degrees thick in declination, with the same visual encoding of depth and overdensity. These images were pre-generated from 509 square images spanning physical sizes from 100 to $1000\,h^{-1}\mathrm{Mpc}$ , with images containing large gaps from survey edges removed (Lamman, 31 Mar 2026).

Users can refresh for a new random image, rotate it, and draw a shape on top of the galaxy wedge. Submitted constellations are stored with the constellation name, the founder’s name, and the location and size of the slice. The gallery can be sorted by “Top discoveries,” “Random,” or “New discoveries,” and users can vote on favorites in the latter two categories, described jokingly as being “under peer review.” Early beta testing involved 20 users, half astronomers, and early users collectively discovered 93 additional constellations (Lamman, 31 Mar 2026). One especially popular example was “Cow Tools,” used to underscore that constellation-making is partly a cultural activity: the patterns are in the sky, but the meaning is made by people.

5. Cosmological homogeneity as an unconventional application

The DESI study frames galactic constellations as an unconventional probe of cosmological homogeneity. Standard analyses of the transition to homogeneity typically use counts-in-spheres, two-point correlation functions, and isotropy tests, and these works place the transition anywhere from about 60 to $500\,h^{-1}\mathrm{Mpc}$ , depending on definition and method (Lamman, 31 Mar 2026). The DESI novelty is to ask at what scales humans still perceive coherent shapes in the galaxy distribution.

For each of the 93 submitted constellations, the study defines size as the maximum distance between any pair of points in the drawing, inferred from the pixel colors in the image and scaled by the total image size to recover a physical span. This allows comparison across slices of different physical extent. The key empirical observation is that all recognizable constellations were smaller than about $60^\circ$ 0, even though the images themselves ranged up to $60^\circ$ 1 across (Lamman, 31 Mar 2026).

The authors interpret this as qualitatively consistent with the cosmological principle and $60^\circ$ 2CDM. Their reading is that the cosmic web becomes too homogeneous for distinctive human-recognizable shapes beyond those scales. At the same time, the paper explicitly states that this is not yet a rigorous cosmological measurement: a stronger test would require a more uniform distribution of available slice sizes, a larger sample, and mock data (Lamman, 31 Mar 2026).

The paper also includes a deliberately tongue-in-cheek statistical claim. Because no $60^\circ$ 3CDM simulation reproduced exactly these constellations, and because no “large fish” appeared in 25 Abacus mock DESI datasets, the absence of Pisces Grandis is jokingly translated into a 96% frequentist “tension.” The paper immediately undercuts that suggestion by stating that such a conclusion demonstrates why frequentist statistics should not be applied to individual structures in cosmological analysis (Lamman, 31 Mar 2026). This passage has become central to the article’s methodological interpretation: galactic constellations are scientifically suggestive but not a precision estimator.

6. Distinction from physical structure in galactic and extragalactic cartography

A recurring misconception is to treat galactic constellations as if they were equivalent to physically defined large-scale structures. The DESI paper explicitly rejects that identification: these shapes are not formally defined physical objects like halos or filaments (Lamman, 31 Mar 2026). That distinction becomes clearer when galactic constellations are compared with tracer-based maps of real Galactic structure.

In Milky Way disc cartography, for example, Gaia DR2 phase-space analysis in guiding coordinates revealed six prominent stellar overdensities out to 5 kpc from the Sun: four associated with the Scutum-Centaurus, Sagittarius, Local, and Perseus spiral arms, and two associated with bar resonances, with corotation at around 6.2 kpc and the outer Lindblad resonance beyond the Solar radius at around 9 kpc (Khoperskov et al., 2019). Those overdensities are identified through phase-space transformation, resampling, unsharp masking, and kinematic diagnostics, and they are interpreted as dynamical imprints of non-axisymmetric Galactic structure rather than as observer-selected shapes.

Similarly, a Gaia DR3 plus 2MASS census of red supergiants uses a young stellar tracer to map the Milky Way’s disk morphology. That work identified 335 new candidate red supergiants and reported a novel population of highly probable red supergiants populating the more distant Scutum-Centaurus arm, in addition to the populous Perseus associations and the Sagittarius group (Messineo, 8 Jun 2025). Here again, the objects of interest are physically motivated overdensities and arm tracers, not visually named figures.

The contrast is instructive. Galactic constellations are a human-centered representation of visually coherent subsets of the cosmic web; spiral arms, bar resonances, and tracer overdensities are astrophysical structures inferred from dynamics, stellar populations, or survey selection functions. This suggests that galactic constellations occupy a distinct epistemic niche: they are neither arbitrary decorations nor replacements for formal structure-finding, but a qualitative interface between cosmological data and human spatial intuition.