Cracking the armor

Crocodile eye

Editor's Introduction

Crocodile head scales are not developmental units but emerge from physical cracking

annotated by

One of a crocodile's most characteristic traits is its thick, scaly skin. Although similar in appearance to elaborately patterned skin of other reptiles, the scales on crocodile heads form irregular polygonal shapes. How are these scales formed? High-resolution imaging and three-dimensional modeling "cracked" the code behind an unexpected developmental process. 

Paper Details

Original title
Crocodile head scales are not developmental units but emerge from physical cracking
Original publication date
Vol. 339 no. 6115 pp. 78-81
Issue name


Various lineages of amniotes display keratinized skin appendages (feathers, hairs, and scales) that differentiate in the embryo from genetically controlled developmental units whose spatial organization is patterned by reaction-diffusion mechanisms (RDMs). We show that, contrary to skin appendages in other amniotes (as well as body scales in crocodiles), face and jaws scales of crocodiles are random polygonal domains of highly keratinized skin, rather than genetically controlled elements, and emerge from a physical self-organizing stochastic process distinct from RDMs: cracking of the developing skin in a stress field. We suggest that the rapid growth of the crocodile embryonic facial and jaw skeleton, combined with the development of a very keratinized skin, generates the mechanical stress that causes cracking.


Amniotes exhibit a keratinized epidermis preventing water loss and skin appendages that play major roles in thermoregulation, photoprotection, camouflage, behavioral display, and defense against predators. Whereas mammals and birds evolved hairs and feathers, respectively, reptiles developed various types of scales. Although their developmental processes share some signaling pathways, it is unclear whether mammalian hairs, avian feathers and feet scales, and reptilian scales are homologous or if some of them evolved convergently (1). In birds and mammals, a reaction-diffusion mechanism (RDM) (2) generates a spatial pattern of placodes that develop and differentiate into follicular organs with a dermal papilla and cycling growth of an elongated keratinized epidermal structure (hairs or feathers) (3). However, scales in reptiles do not form true follicles and might not develop from placodes (4). Instead, reptilian scales originate in the embryo from regular dermo-epidermal elevations (1). Whereas the regular spatial organization of scales on the largest portion of the reptilian body is determined by a RDM, additional positional cues are likely involved in the development of the scale plates present on the head of many snakes and lizards. These head scales form a predictable symmetrical pattern (Fig. 1A) and provide mechanical protection.


Fig. 1.  Spatial distribution of head scales. (A) Head scales in most snakes (here, a corn snake) are polygons (two upper panels) with stereotyped spatial distribution (two lower panels): left (yellow) and right (red) scale edges overlap when reflected across the sagittal plane (blue). (B) Polygonal head scales in crocodiles have a largely random spatial distribution without symmetrical correspondence between left and right. (C) Head scales from different individuals have different distributions of scales’ sizes and localizations (blue and red edges from top and bottom crocodiles, respectively).

Method: 3D geometry and color-texture reconstruction

The authors took 120 color pictures of each animal to create detailed, three-dimensional models of reptile heads. Watch this video in which the authors further explain their modeling methods.

Panel A

Panel A shows the spatial distribution of head scales of the corn snake. The red and yellow lines represent scale edges on the left and right side of the head. The bottom panel shows that the scale pattern is symmetrical across the head.

Panel B  

Panel B shows the spatial distribution of head scales of the Nile crocodile. As shown in the bottom panel, the scale patterns on the left and right side of the crocodile head are not symmetrical to each other, in contrast to pattern of the corn snake.

Panel C

Panel C also shows the spatial distribution of the crocodile head scales. In this panel, the scale patterns of two individual crocodile heads are shown. By overlaying the scale patterns, the authors show that the scale size and location are not consistent between individuals. The exact scale pattern of an individual is unique and could be used, as fingerprints are in humans, to identify individuals.

The face and jaws of crocodilians are covered by polygonal scales (hereafter called “head scales”) that are strictly adjoining and nonoverlapping, but these polygons are irregular and their spatial distribution seems largely random (Fig. 1, B and C). Using high-resolution three-dimensional (3D) geometry and texture reconstructions (57), as well as developmental biology techniques, we show that crocodilians’ head scales are not genetically controlled developmental units and that their spatial patterning is generated through physical cracking of a living tissue in a stress field. This phenomenon might not involve any specific genetic instruction besides those associated with cell proliferation and general physical parameters such as skin stiffness and thickness.

By marking and analyzing various features directly on 3D models of multiple Nile crocodile (Crocodylus niloticus) individuals (Fig. 1 and movie S1), we show that spatial distribution of head scales is largely random.

First, reflection of the network of scales’ edges across the sagittal plane indicates high variability between the left and right head pattern (Fig. 1B and fig. S1A). Second, nonrigid alignment (8) of head geometries from different individuals indicates a similarly large variability in scale patterns in terms of polygons’ sizes and localizations (Fig. 1C and fig. S1B).

This combination of order and chaos in the distribution of head scales is reminiscent of the topological assemblage of soap foams (910). Recent studies used the 2D foam model for studying self-organizing principles and stochastic processes shaping epithelial topology during growth and homeostasis (11–13) because the causal cell-surface mechanics is comparable to the physics of foam formation (14). Similarly, the pattern of crocodile head scales could result from energy minimization of contact surfaces among genetically determined elements (scales). However, two other mechanisms could generate random distributions of polygonal elements: (i) a RDM patterning the spatial organization of genetically determined developmental units, as for mammalian hairs or avian feathers, and (ii) cracking of a material layer causing its fracture into adjacent polygonal domains (15).

Although stochastic patterns generated by these processes share some universal mathematical properties (see supplementary materials), foams and crack patterns are generated by very different physical phenomena that may be identified on the basis of other statistical features. First, crocodile head scales do not show a good fit to the area distribution function expected for foams (fig. S3). Second, a fundamental difference between foams and crack patterns is that the latter can exhibit incomplete edges (15), of which many are observed on the head of crocodiles (Fig. 2A).


Fig. 2.  Signatures of cracking. (A) Many scale edges on crocodiles’ head are unconnected at one or both ends. (B) Three incomplete cracks interact symmetrically. (C) Edges reorienting (arrows) and connecting with angles close to 90°. (D) “Laddering” between parallel primary cracks. (E) 90° network border connections. (F) DPRs are pigmented sensory organs (left) with a modified epidermis (right, section at embryonic stage E70, that is, at 70 days of egg incubation) and a pocket of branched nerves (white arrowheads). (G) Incomplete cracks stopping in close proximity to a DPR (orange dots). (H) Crack propagation avoids DPRs.

Panels A and B

Panels A and B show incomplete scale edges, that is, edge ends that do not connect to another edge. The model based on soap foam formation does not predict the formation of incomplete edges, but a model of cracking due to shrinkage does. Therefore, the authors conclude that these observations support the hypothesis that scales form by cracking.

Panel C

Panel C shows a hierarchical scale edge pattern. A crack propagating from a preexisting crack end tends to form at a 90 degree angle with the second edge it will connect to. The black arrows in Panel C indicate examples of propagating cracks. Panels D and E show ‘laddering’ cracks, indicated by parallel cracks that connect to two older cracks or to an external edge.

Panel F  

Panel F is a histological stain of the epidermis from a crocodile embryo. The dome pressure receptors, i.e., integumentary sensory organs found on the crocodile head and jaw, appear as brown spots. White arrows indicate nerves branching below the dome pressure receptor.

Panes G and H 

Panels G and H show the spatial relationship between dome pressure receptors, indicated as orange dots, and scale edges. Edges end at the dome pressure receptors or avoid them.

Another key feature is the angle among edges at nodes. In foams, edges are circular arcs intersecting only three at a time with an angle of 120°, as imposed by the three instantaneous equal length-tension force vectors acting at a node. This rule is observed in all types of foams, including animal epithelia (1216), although the distribution of angles can be widened due to local stress generated by cell division and growth. On the other hand, crack patterns can generate various angle distributions. Nonhierarchical cracking arises when fractures propagate simultaneously (Fig. 2B), and junctions tend to form at 120° (1718). Furthermore, when a crack front splits, or when multiple cracks are nucleated from a single point, the junctions among edges also tend to be 120°. Conversely, crack patterns can be hierarchical (1719); that is, fractures are formed successively, and propagating cracks will tend to join previous cracks at a 90° angle. Indeed, the local stress perpendicular to a crack is relaxed and concentrates at the tip of the crack (explaining its propagation), but the stress component parallel to the crack is not affected. Hence, as cracks propagate perpendicularly to the direction of the maximum stress component, a secondary crack can turn when it approaches an older one and tends to join it at 90°. Similarly, if a crack starts on the side of an older crack, it will initially tend to propagate at a right angle (17). Multiple examples of 90° connections and incomplete edges reorienting their propagation front are visible on the crocodiles face and jaws (Fig. 2C). We also observe “laddering” patterns (17) of paired parallel primary fractures with perpendicular multiple secondary cracks (Fig. 2D) and internal edges connecting perpendicularly to the border of the network (Fig. 2E). The distribution function of edge angles is bimodal in many crocodiles analyzed (fig. S4A), suggesting either that hierarchical and nonhierarchical cracking processes coexist or that head scale networks undergo a “maturation” process (20–22) (see supplementary text).

Dome pressure receptors (DPRs) are pigmented round submillimetric sensory organs (Fig. 2F), distributed on the crocodile face and jaws, that detect surface pressure waves, allowing crocodiles to swiftly orient, even in darkness, toward a prey perturbing the water-air interface (23). The dome shape of DPRs is due to a modified epidermis and the presence of a pocket of various cell types in the outermost portion of the dermis (Fig. 2F). We marked the localization of DPRs on the 3D models of all scanned individuals (orange dots, Fig. 2, G and H). Many of the cracks that have stopped their course did so close to a DPR (Fig. 2G and fig. S4C). Given that the most frequent cause for fracture arrest is when the crack front meets a heterogeneity in the system (15), it is likely that the modified skin thickness and composition at and around DPRs constitute such heterogeneities. In addition, the course of many edges avoids DPRs (Fig. 2H and fig. S4C).

The overall distribution of DPRs seems rather homogeneous except where the density is increased near the teeth and decreased at the back of the jaws and on the top of the face (fig. S5). Different crocodile individuals differ by as much as 21% and 48% in their total number of DPRs and crack edges, respectively. Remarkably, these two interindividual variations are inversely correlated: Crocodiles with fewer DPRs have more crack edges (fig. S4D). Given that the development of DPRs precedes cracking, this correlation suggests that DPRs constrain cracking, as already implied by Fig. 2, G and H. Despite the fact that the distributions of cracks and DPRs both have a strong stochastic component, the constraining effect of DPRs on cracking is noticeable: The edges tend to travel along the zones of DPRs lowest local density (fig. S4E).

The archetypal cracking process in physics is due to shrinkage [through removal of a diffusing quantity, either heat or a liquid (20)] of a material layer adherent to a nonshrinking substrate (15, 17), such that a stress field builds up and causes fractures when the stress exceeds a threshold characteristic of the material. Crocodiles have a particularly thick and rigid skin due to the presence of a highly collagenous dermis and an epidermis rich in β-keratins (24). The skin covering their head shows a yet thicker (about 2×) and more keratinized epidermis. We suggest that the rapid growth of the crocodile embryonic facial and jaw skeleton (relative to the size of the neurocranium), combined with the development of a very keratinized skin, generates the mechanical stress that causes cracking. Here, it is not the cracking layer that shrinks but the underlying substrate layer that grows. It explains that first-order cracks (fig. S6) tend to traverse the width of the face because the head is growing longitudinally faster than in other directions.

In snakes and lizards, scales are developmental units: Each scale differentiates and grows from a primordium that can be identified by in situ hybridization with probes targeting genes belonging to signaling pathways involved in early skin appendage development (14). The large head scales form a predictable pattern following positional cues, such that the identity of adult snake head scales can be recognized while they develop from primordia in the embryo (Fig. 3A). In crocodiles, all postcranial scales follow that same principle of development (Fig. 3B): Spatial distribution of primordia is established, then each primordium differentiates, first into a symmetrical elevation and second as an oriented asymmetrical scale overlapping with more posterior scales (Fig. 3C).


Fig. 3.  Crocodile head scales are not developmental units. (A) In snakes, each body scale (ventral, v; latero-dorsal, ld) differentiates from a primordium (Shh gene probe for in situ hybridization, corn snake embryo); head scales also develop from primordia, with positional cues determining scale identity (la, labial scales; r, rostral; n, nasal; in, internasal; pf, prefrontal; pro, preocular; so, supraocular; pto, postocular). (B) Postcranial scales (zoom on trunk, Ctnnb1 probe) also develop from primordia that (C) differentiate into symmetrical, then oriented asymmetrical and overlapping, scales. (D) Crocodile head scales never form scale primordia [nor developmental stages shown in (C)] but, instead, develop a pattern of DPRs (one DPR circled with dotted line; dome shape visible at E45) before any scale appears (probe: Ctnnb1).

Scale primordia express specific genes

The authors use in situ hybridization to measure expression of these specific genes to identify primordia. In situ hybridization is a technique in which a labeled piece of RNA or DNA, called a probe, complementary to the gene of interest, is localized to a tissue. The labeled probe will base pair with the mRNA sequence of interest, and thus label cells wherever a transcript of the gene of interest is present. To learn more about in situ hybridization, watch this video.

Panel A

In Panel A, the authors perform an in situ hybridization on a corn snake embryo. The probe used is complementary to a transcript expressed in scale primordia, and therefore the dark spots indicate scale primordia. The adult corn snake’s head scales are shown in the right panel in a variety of colors. By comparing the position of the scale primordia to the adult head scales, the authors conclude that the corn snake’s head scales develop from scale primordia.  

Panel C  

Next, the authors examined the crocodile postcranial scales by performing an in situ hybridization on a crocodile embryo.  As in Panel A in the snake embryo, the in situ probe used is complementary to a transcript expressed in scale primordia, so the dark spots indicate scale primordia.  Panel C shows, in cross section, the time-course of morphological transformations leading to scale formation: elevations then overlapping of the scales. The authors conclude that crocodile postcranial scales also develop from scale primordia.

Panes G and H 

The authors then examine crocodile head scales. Panel D shows an in situ hybridization of the crocodile head and jaw, using the same probe as in Panel B. Tiny dark spots appear on crocodile head and jaw, but they correspond to the DPR/ISO receptors, not the scales. Hence, the head and jaw scales do not originate from scale primordial, but DPR/ISOs do. The histological sections in the right side of the panel confirm that the formation of dome pressure receptors occurs before the formation of head scales.

However, crocodile head scales do not form from scale primordia or further developmental stages. Instead, a pattern of DPRs primordia is generated on the face and jaws: The dome shape of DPRs has already started to form before any scale appears (Fig. 3D). Afterward, grooves progressively appear, propagate, and interconnect (while avoiding DPRs) to form a continuous network across the developing skin (Fig. 4A). The process generates polygonal domains of skin, each containing a random number of DPRs. Therefore, scales on the face and jaws of crocodiles (i) are not serial homologs of scales elsewhere on the body and (ii) are not even genetically controlled developmental units. Instead, they emerge from physical cracking.


Fig. 4. Crocodile head skin cracks during development. (A) There is no sign of cracking at E45 (but DPRs primordia are already developed, Fig. 3D), then primary cracks (arrowheads) appear on the sides of the upper jaws and progress toward the top of the face (dotted line). At E65, primary cracks reached the top of the head and are followed by secondary cracks in other orientations (arrows). (B) Three sequential skin sections along primary (pc) and secondary (sc) cracks (ep, epidermis; de, dermis; bo, bone tissue). (C) Antibody to pan cadherin stains the whole epidermis, antibody to proliferating cell nuclear antigen (PCNA) indicates increased proliferation (arrows), and terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) assay indicates absence of apoptosis in cracks.

Panel A

Panel A shows the development of crocodile head skin cracks as the embryo develops. Figure 3D shows the development of dome pressure receptors when the embryos is 45 days old, and the left panel of Figure 4A shows that no cracks have formed at that stage. Cracks begin to develop when the embryo is 50 to 55 days old.

Panel B  

Panel B is a histological cross section along a crack on the crocodile embryonic skin. From left to right, the panels show sequential sections of the skin. The crack extends from the top of the epidermis into the dermis but not into the bone tissue.


In Panel C, the authors use immunofluorescence to understand where specific proteins are localized. Immunofluorescence is a technique for visualizing protein localization. To learn more about immunofluorescence, click here.

In addition to standard immunofluorescence, the authors use a terminal deoxynucleotidyl transferase-mediated deoxyuridine triphosphate nick end labeling (TUNEL) assay to identify cells that have undergone apoptosis. Apoptotic cells have damaged DNA, and that damage includes nicks that can be repaired by an enzyme that inserts a deoxyuridine triphosphate (dUTP) at the site of the nick. When enzyme and dUTPs that are labeled (in this case in red) are added to cells, only apoptotic cells will incorporate the dUTPs and appear red.  

Panel C

In Panel C, the authors examine cell proliferation in a cross section of the embryonic epidermis. In the first and third panel, a protein expressed in all epidermal cells is labeled, and therefore all epidermal cells appear green.  In the second and fourth panel, a protein found in proliferating cell is labeled, and therefore proliferating cells appear green. The second and fourth panel show that cells within the cracks are proliferating. The second and fourth panel also show, using the TUNEL assay, that apoptosis is reduced within the cracks.

During a typical cracking process, fractures are nucleated at the upper surface but quickly spread downward and affect the whole thickness of the material layer (19). The developing skin on the crocodile’s head similarly reacts to the stress field as it develops deep groves that can reach the stiff underlying tissues (Fig. 4B). Our analyses indicate that cell proliferation in the epidermis layer is vastly increased in the deepest region of the skin grooves corresponding to cracks (Fig. 4C), suggesting that a healing process allows the skin layer to maintain its continuous covering. The local biological process (cell proliferation) might be driven by the purely physical parameter (mechanical stress) as follows: In zones of highest stress, local bulging is nucleated. The local stress component perpendicular to the bulge is relaxed and concentrates at its tip, explaining the propagation of both the stress and proliferation maxima (hence, the corresponding propagation of the bulge). In a manner entirely analogous to true physical cracking, the bulge front would propagate perpendicularly to the direction of maximum stress components, explaining the topology of the resulting random polygonal domains of skin. The role of proliferation reinforces the above suggestion that crack patterns in crocodiles might experience maturation (2022), explaining the observed mixture of hierarchical and nonhierarchical features (see supplementary text and fig. S4A).

We have shown that the irregular polygonal domains of skin on the crocodile face and jaws are produced by cracking, a mechanism that is distinct from those generating scales on the postcranial portion of the crocodile body, as well as on the body and head of all other reptiles. This cracking process is primarily physical. However, it does not mean that genetically controlled parameters are irrelevant. For example, although a crack pattern is visible in all crocodilian species, spatial distribution varies considerably, possibly because of species-specific skull geometry and growth but also skin composition and thickness. Given that these parameters, as well as cell proliferation, are genetically controlled, the variation of head crack patterns among crocodilian species is likely due to an interplay between physically and genetically controlled parameters. Our study suggests that, besides RDM, a larger set of physical self-organizational processes contribute to the production of the enormous diversity of patterns observed in living systems.

Supplementary Materials

Material and Methods

Supplementary Text

Figs. S1 to S6

Table S1

Movie S1

References (2533)

References and Notes 

  1. C. Chang et al., Reptile scale paradigm: Evo-Devo, pattern formation and regeneration. Int. J. Dev. Biol. 53, 813 (2009).

  2. M. Turing, The chemical basis of morphogenesis. Philos. Trans. R. Soc. London Ser. B 237, 37 (1952).

  3. M. R. Schneider, R. Schmidt-Ullrich, R. Paus, The hair follicle as a dynamic miniorgan. Curr. Biol. 19, R132 (2009).

  4. D. Dhouailly, A new scenario for the evolutionary origin of hair, feather, and avian scales. J. Anat. 214, 587 (2009).

  5. K. N. Snavely, S. M. Seitz, R. Szeliski, Photo tourism: Exploring photo collections in 3D. ACM Transactions on Graphics (Proc. SIGGRAPH 2006) 25, 835 (2006).

  6. Y. Furukawa, J. Ponce, Accurate camera calibration from multi-view stereo and bundle adjustment. Int. J. Comput. Vis. 84, 257 (2009).

  7. M. Kazhdan, M. Bolitho, H. Hoppe, Eurographics Symposium on Geometry Processing 2006, 61 (2006).

  8. H. Li, B. Adams, L. J. Guibas, M. Pauly, ACM Transactions on Graphics (Proc. SIGGRAPH Asia 2009) 28, 175 (2009).

  9. D. Weaire, S. Hutzler, The Physics of Foams (Oxford Univ. Press, Oxford, 1999).

  10. D. Weaire, N. Rivier, Soap, cells and statistics: Random patterns in two dimensions. Contemp. Phys. 25, 59 (1984).

  11. T. Lecuit, P. F. Lenne, Cell surface mechanics and the control of cell shape, tissue patterns and morphogenesis. Nat. Rev. Mol. Cell Biol. 8, 633 (2007).

  12. M. C. Gibson, A. B. Patel, R. Nagpal, N. Perrimon, The emergence of geometric order in proliferating metazoan epithelia. Nature 442, 1038 (2006).

  13. D. A. W. Thompson, On Growth and Form (Cambridge Univ. Press, Cambridge, 1917).

  14. R. Farhadifar, J. C. Röper, B. Aigouy, S. Eaton, F. Jülicher, The influence of cell mechanics, cell-cell interactions, and proliferation on epithelial packing. Curr. Biol. 17, 2095 (2007).

  15. H. J. Herrmann, S. P. Roux, Eds.,Statistical Models for the Fracture of Disordered Media: Random Materials and Processes (North-Holland, Elsevier, Amsterdam, 1990).

  16. T. Hayashi, R. W. Carthew, Surface mechanics mediate pattern formation in the developing retina. Nature 431, 647 (2004).

  17. K. A. Shorlin, J. R. de Bruyn, M. Graham, S. W. Morris, Development and geometry of isotropic and directional shrinkage-crack patterns. Phys. Rev. E Stat. Phys. Plasmas Fluids Relat. Interdiscip. Topics 61, 6950 (2000).

  18. E. A. Jagla, A. G. Rojo, Sequential fragmentation: The origin of columnar quasihexagonal patterns. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 65, 026203 (2002).

  19. S. Bohn, J. Platkiewicz, B. Andreotti, M. Adda-Bedia, Y. Couder, Hierarchical crack pattern as formed by successive domain divisions. II. From disordered to deterministic behavior. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 71, 046215 (2005).

  20. L. Goehring, L. Mahadevan, S. W. Morris, Nonequilibrium scale selection mechanism for columnar jointing. Proc. Natl. Acad. Sci. U.S.A. 106, 387 (2009).

  21. L. Goehring, S. W. Morris, Order and disorder in columnar joints. Europhys. Lett. 69, 739 (2005).

  22. L. Goehring, R. Conroy, A. Akhter, W. J. Clegg, A. F. Routh, Evolution of mud-crack patterns during repeated drying cycles. Soft Matter 6, 3562 (2010).

  23. D. Soares, Neurology: An ancient sensory organ in crocodilians. Nature 417, 241 (2002).

  24. L. Alibardi, L. W. Knapp, R. H. Sawyer, Beta-keratin localization in developing alligator scales and feathers in relation to the development and evolution of feathers. J. Submicrosc. Cytol. Pathol. 38, 175 (2006).

  25. K. N. Snavely, Scene Reconstruction and Visualization from Internet Photo Collections. PhD thesis, University of Washington (2008).

  26. Y. Furukawa, B. Curless, S. M. Seitz, R. Szeliski, Towards Internet-scale Multi-view Stereo. IEEE Conference on Computer Vision and Pattern Recognition 2010, 1434 (2010).

  27. V.C.L.I., CNR, MeshLab v1.3.0a; (2011).

  28. H. Li, BeNTO3D, Pre-release developmental version; (2012).

  29. J. F. Sadoc, R. Mosseri, Frustration Géométrique. A. Saclay, Ed. (Eyrolles, Paris, 1997).

  30. N. Di-Poïet al., Changes in Hox genes’ structure and function during the evolution of the squamate body plan. Nature 464, 99 (2010).

  31. S. Bohn, L. Pauchard, Y. Couder, Hierarchical crack pattern as formed by successive domain divisions. I. Temporal and geometrical hierarchy. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 71, 046214 (2005).

  32. E. A. Jagla, Maturation of crack patterns. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 69, 056212 (2004).

  33. K. J. Stine, S. A. Rauseo, B. G. Moore, J. A. Wise, C. M. Knobler, Evolution of foam structures in Langmuir monolayers of pentadecanoic acid. Phys. Rev. A 41, 6884 (1990).

Acknowledgments: This work was supported by the University of Geneva, the Swiss National Science Foundation, and the Schmidheiny Foundation. A. Tzika helped with in situs. H. Li assisted with nonrigid registration. We thank R. Pellet for assistance in mechanics design and A. Roux, M. Gonzalez-Gaitan, B. Chopard, U. Schibler, and anonymous reviewers for useful comments and suggestions.