Introduction
The study of the behavior of ancient organisms is elusive, because trace fossils are the only tangible record of animal behavior. Trace fossils provide a geologic record with a potential of showing changes in animal behavior, that is, behavioral evolution, throughout geologic time (Seilacher Reference Seilacher1974). Previous workers often have documented the record of behavioral evolution through time simply by counting the numbers of ichnogenera and ichnospecies as they occur throughout the fossil record and by adding up those numbers in each time period (Seilacher Reference Seilacher1974; Uchman Reference Uchman2003, Reference Uchman2004). Other behavioral evolution studies include describing the evolution of the trace fossil Zoophycos through time (Seilacher Reference Seilacher1986; Bottjer et al. Reference Bottjer, Droser and Jablonski1988; Olivero Reference Olivero2003) and using Lévy random walks as a search strategy to help explain the evolution of foraging behavior (Sims et al. Reference Sims, Reynolds, Humphries, Southall, Wearmouth, Metcalfe and Twitchett2014). A different method for cataloging change among related groups is to measure those groups based on quantitative morphological characteristics. As with morphological cladistics, this approach removes the role of taxonomic nomenclature, which does not have the ability to capture the full range of morphological possibilities, while still analyzing meaningful aspects of the forms of the traces.
A major problem with deciphering behavioral evolution from trace fossils is that the same animal can make many different types of trace fossils, and many different types of animals sometimes can make the same trace fossil. Trace fossils produced in a restrictive environment with a similar mode of preservation should be used to reduce the number of trace maker variables. For that reason, graphoglyptid trace fossils and other trace fossils preserved in the same environment were used in this study.
Graphoglyptids are predepositional, geometrically complex, predominantly horizontal, open-burrow systems commonly preserved in convex hyporelief on the soles of deep-sea turbidite beds. Graphoglyptids are instructive trace fossils for studying behavioral evolution, because they are typically found in deep-sea deposits, where they formed as open tunnels within a few centimeters of the sediment surface, and they typically are preserved as casts on the soles of turbidite beds. The intricate geometric patterns exhibited by the majority of graphoglyptid ichnogenera display a similar degree and type of complexity to warrant the assumption that they may have been made by closely related species (Seilacher Reference Seilacher1977); however, there is some debate about that issue (Miller Reference Miller2012).
Basis of Analyses
This study analyzed graphoglyptids and a few other related trace fossils preserved as turbidite casts (i.e., Gordia and Helminthopsis). Restricting the study to graphoglyptid trace fossils limits the number of possible variables, because graphoglyptids generally were created in similar environments with similar conditions and were preserved under similar circumstances. The analyses in this study focused only on those forms preserved in convex hyporelief on the sole of turbidite beds, except in rare instances where other preservation modes were found, for instance, in modern graphoglyptids (Ekdale Reference Ekdale1980). The analyses were also limited to graphoglyptids that were formed predominantly in a horizontal plane (e.g., Cosmorhaphe and Paleodictyon), because graphoglyptids that are initially formed three dimensionally (e.g., Lorenzina, Desmograpton, and Helicolithus) do not have enough material preserved in individual specimens to glean much meaningful behavioral information. It is only by looking at multiple examples that it is possible to piece together the complete forms in three dimensions (Uchman Reference Uchman1998).
Topology Groups
Topology involves describing something in terms of specific geometric characteristics. For this study’s purposes, the graphoglyptid trace fossils were grouped into four different topological categories: meandering forms, spiraling forms, branching forms, and network forms (Fig. 1; also see Lehane and Ekdale Reference Lehane and Ekdale2014). Each trace fossil was analyzed and compared with all of the other trace fossils within each of the four topological groups.
Figure 1 Topological categories of graphoglyptid trace fossils. A, Meandering forms. B, Spiraling forms. C, Branching forms. D, Network forms.
Meandering Forms
Meandering forms (Fig. 1A) are continuous nonbranching tunnels that have no natural breaks in them (barring effects of erosion or imperfect preservation) and an overall sinuous morphology. Many of these trace fossils likely have branches that extend to the sediment–water interface; however, the majority of the trace is preserved in two dimensions as a continuous meander and therefore is included in this category. The trace fossils within this category are Belocosmorhaphe, Belorhaphe, Cosmorhaphe, Helminthopsis, Helminthorhaphe, Oscillorhaphe, Paleomeandron, and Spirocosmorhaphe. Although Helminthopsis is not strictly a graphoglyptid (it does not feature the intricate patterns of graphoglyptids), it is often found as a hypichnial ridge on the base of turbidites. This indicates that these forms of Helminthopsis had a similar construction style to graphoglyptids (open burrow), and the trace maker lived in a similar environment (deep sea). For these reasons, Helminthopsis, specifically only those forms found as hypichnial ridges on deep-sea turbidites, have been included in this study as an example of primitive meandering behavior.
Spiraling Forms
Spiraling forms (Fig. 1B) are continuous nonbranching tunnels that have no natural breaks in them (barring effects of erosion or imperfect preservation) and an overall spiraling morphology. Only one graphoglyptid ichnogenus fits into this category, Spirorhaphe, which ranges from the Ordovician to the modern. Although this is a long time range, aside from one specimen in the Ordovician and two in the modern, the remaining 24 specimens all occur during a 36 Myr time span from the Late Cretaceous to the Eocene. For this reason, no significant conclusions about the behavioral evolution of Spirorhaphe over the course of the Phanerozoic can be reached at this time, so this ichnogenus will not be discussed further in this paper.
Branching Forms
Branching forms (Fig. 1C) are similar to the meandering forms, except that a given burrow splits into two or more burrows repeatedly. These branches do not reconnect to the main trunk or any other branch. The graphoglyptids that are included in this category are Protopaleodictyon, Ubinia, and Urohelminthoida.
Network Forms
Network forms (Fig. 1D) are similar to branching forms, except the branches reconnect to each other or reconnect back to the main trunk. To be assigned to this group, a sample must exhibit at least one complete cell (an individual unit within a mesh or network). The trace fossils included in this category are Gordia, Megagrapton, Paleodictyon, and Squamodictyon. Although Gordia is not a graphoglyptid by definition (it does not contain the intricate patterns associated with graphoglyptids), there are many samples of Gordia that have been found preserved as raised ridges (hypichnally) on the base of turbidites. Gordia, specifically only those examples found as hypichnial ridges on the base of turbidites, is included in the network analyses as a primitive example of network behavior.
Materials
For these analyses, graphoglyptid images were taken from field photographs, museum specimens, and illustrations in published literature. The field photographs were taken in the Guipúzcoan Flysch, Higuer–Getaria Formation (Ypresian, lower Eocene), Zumaia, Spain, which is known as one of the best locations in the world to find graphoglyptids; in the Vera Basin (Messinian, Miocene), Almería, southeastern Spain; and among the turbidites at Point Saint George within the Franciscan Complex (Early Cretaceous), northern California. The museum specimens that were photographed are housed in the University of Utah Ichnology Collection, Salt Lake City, Utah; the Institute of Geological Sciences, Jagiellonian University, Krakow, Poland; and the University of California Museum of Paleontology, Berkeley, California. The remaining samples were collected from previously published images in the literature. The source of many of the references consulted was the list of reference citations in Uchman (Reference Uchman2004), which is one of the most comprehensive collections of turbidite trace fossil papers known. A list of known turbidite deposits is included within (Supplementary File 1), which highlights how many turbidite deposits contain graphoglyptids and how many of those were used in these analyses.
More than 1000 samples were analyzed for potential inclusion in the study. The samples were analyzed for completeness, diversity, and time periods. Specimens that were more complete were chosen over fragmented specimens in order to get a better understanding of the complete trace fossil shapes. Samples were also chosen so that the authors could feel certain that a full range of morphological values were accounted for during any particular time period. Although there is not a lot of data from Paleozoic occurrences, we are confident that a complete range of possible data was represented in the analyses. Literature searches focused on time periods that were not represented in field studies or museum samples. A total of 387 samples were chosen for analysis. These samples range from the early Cambrian to modern examples (Fig. 2; Supplementary Files 1 and 2) and primarily from sites in Europe and North America with some samples representing other parts of the world (Supplementary File 3). This total includes 140 meandering forms, 65 branching forms, and 182 network forms.
Figure 2 Geologic ranges of the graphoglyptid taxa analyzed. M, meandering forms; S, spiraling forms; B, branching forms; N, network forms.
Methodology
The morphology of the trace fossils were studied analytically to describe the evolutionary trends of the studied graphoglyptids through time. Due to their preservation, graphoglyptid trace fossils are regarded as planar structures, and therefore they allow for two-dimensional analyses of morphology. A chart of the analytical results is provided (Supplementary File 4). Most of the analytical techniques were described in detail in Lehane and Ekdale (Reference Lehane and Ekdale2013a, Reference Lehane and Ekdale2014), and further information on each technique may be obtained there; however, some changes or extensions of those analyses did occur, which are described herein (Supplementary File 5).
Each trace fossil was photographed and traced following the midline of the burrow (Supplementary File 2). The basic measurements of the trace fossils included the meander size, also known as the motility index (MI), for meandering and branching forms, and the mesh size (MS), the average cell size of a network form. The average branching angles (BAs) were measured for both the branching and the network forms.
For the meandering and branching forms, the tortuosity (τ) was calculated by dividing the average length of the line by the straight-line distance between the endpoints of the trace (Fig. 3A). These measures were compared using similarly sized sections in each trace fossil (5 cm, 10 cm, 15 cm, etc.). For instance, equal-length segment averages could be compared between different trace fossils. The τ of the network forms was determined in a similar way, except the τ values for all of the available pathways were measured and averaged together (Fig. 3B). This was calculated from an updated Matlab script (Supplementary File 6) that was employed in previous analyses (Lehane and Ekdale Reference Lehane and Ekdale2014).
Figure 3 Tortuosity measurements. A, Tortuosity measurements for meandering forms. a and c are endpoints, b is the midpoint. L1, L2, and L3 represent the lengths of each respective line. B, Tortuosity measurement for network forms. d, e, f, and g are the endpoints of the measuring lines. L4 and L5 represent the lengths of each respective line.
Some analyses were performed on all of the topological groups. These methods include the calculation of the fractal dimension, which is the amount of space the trace fossil covers combined with the intricacy of the design into one number (Lehane and Ekdale Reference Lehane and Ekdale2013a). The next analysis involved calculation of the occupied space percentage, which is the amount of the surface that is covered by the trace fossil. The final analysis involved was the calculation of the burrow area shape (BAS), which was used to determine the fitness of the fossil for study. Any trace fossil that had too many gaps within the form would produce a low BAS and was therefore omitted from a full analysis.
Results
Not all of the analyses are discussed in detail; however, all of the results are presented in Supplementary File 4. Within the figures, the median value of the results for each time period are plotted with the error bars represented by the 95% confidence interval (Gardner and Altman Reference Gardner and Altman1986). The 95% confidence interval provides the range of morphological values while removing the outliers and reducing the clutter of plotting all of the available data.
Meandering Forms
The study set of the meandering forms consisted of 140 specimens, ranging from the Cambrian to the modern, including the ichnogenera Belocosmorhaphe, Belorhaphe, Cosmorhaphe, Helminthopsis, Helminthorhaphe, Oscillorhaphe, Paleomeandron, and Spirocosmorhaphe (Fig. 2). The measures of burrow width (W), fractal dimension (DBox), τ, and MI were plotted against the age of the specimens to determine any noticeable trends (Figs. 4 and 5). Within the meandering forms, there is a large time gap in data from the Devonian to the Early Cretaceous. Part of the cause of this gap is the 95% confidence interval, which does not allow for the plotting of time periods with only one specimen or a group of specimens for which the specific analysis can be applied to only one of them (e.g., in the Jurassic and the Carboniferous).
Figure 4 Results of analyses of meandering forms. A, Measure of the burrow width (W) in mm over time, plotted in millions of years before present. B, Measure of the fractal dimension calculated by the box method (DBox) over time, plotted in millions of years before present. Black solid shapes and error bars for both plots represent the 95% confidence interval of the data set per time period, and the ages represent the median age of the estimated age range of the specimen. Light gray shapes represent the raw data. Geologic timescale provided for clarity. The darker gray background and arrows are included to help illustrate these changes in values.
Figure 5 Results of analyses of meandering and branching forms. A, Measure of the τ of meandering forms over time, plotted in millions of years before present. The 5 cm, 20 cm, and 30 cm τ are plotted on the graph. B, Measure of the MI of meandering forms (triangles) and branching forms (circles) over time, plotted in millions of years before present. Black solid shapes and error bars for both plots represent the 95% confidence interval of the data set per time period, and the ages represent the median age of the estimated age range of the specimen. Light gray shapes represent the raw data. Geologic timescale provided for clarity. The darker gray background and arrows are included to help illustrate these changes in values.
The plot of W over geologic time (Fig. 4A) shows two groupings of meandering traces: (1) Cambrian to Devonian and (2) Early Cretaceous to modern. The two groups show distinctly different patterns, with the W of the Paleozoic specimens having a very large range of about 8 mm (95% confidence interval of ~0.2 to ~7.8 mm), while the Mesozoic and Cenozoic group has a range of only about 4 mm (95% confidence interval of ~0.1 to 4.0 mm). The Paleozoic samples show a lack of similarity between the trace fossils within and across the different time periods, while the later samples show much less variation between and within each time period.
The plot of DBox over geologic time (Fig. 4B) shows two distinct patterns of the meandering traces. Up until the Late Cretaceous (the dashed line on Fig. 4B), the DBox was not consistent, and the points occurred over a large range of values. After the Late Cretaceous, the DBox values were close to their maximum and proceeded to steadily decline after that.
Trends of τ through geologic time (Fig. 5A) are a little more complex than the previous analyses. The graph shows the median values and 95% confidence intervals for only the 5 cm, 20 cm, and 30 cm τ calculations to avoid making the graph unreadable. Even though there is a large gap in time between the two groups of trace fossils, there is a trend that can be seen going from the 5 cm τ up to the 30 cm τ. Through time, the 5 cm τ has remained basically the same. The 20 cm τ shows an increasing trend from the Paleozoic through the Cenozoic with a sudden drop during the Oligocene. The 30 cm τ continues this trend with an increase from the Paleozoic through the Cenozoic, with a drop-off in the Oligocene as well. By looking at just the average values from the 5 cm τ through the 30 cm τ, it can be seen that the longer the traces are, the more tortuous they tend to be.
The MI for both the meandering and the branching forms is plotted over geologic time (Fig. 5B). Even though many of the same analyses can be applied among the different topological forms, the trends that they display are often different. However, in this instance, the meandering and branching forms exhibit similar patterns, so they have been plotted on the same graph. Several trends for the MI can be discerned from this graph. The first is that there is a drop in MI from the Cambrian through the rest of the Paleozoic. The low MI also matches the τ for the Cambrian or the rest of the Paleozoic. The low MI values then start to rise, peaking in the Late Cretaceous. The MI remains fairly high until the Oligocene, where the MI value drops off steeply to almost Paleozoic values, and then it increases again in the Miocene.
Branching Forms
The study set of branching forms consists of 65 specimens, ranging from Ordovician to Miocene in age, including the ichnogenera Protopaleodictyon, Ubinia, and Urohelminthoida (Fig. 2). The measures of MI, DBox, W, and BA were plotted against the age of the specimens to determine any noticeable trends (Figs. 5B and 6). The branching form can be considered a cross between the meandering forms and the network forms, so the individual measurements that are the same as the other two forms were compared directly with them (i.e., the MI with the meandering forms and the BA with the network forms).
Figure 6 Results of analyses of branching forms. A, Measure of the fractal dimension calculated by the box method (DBox) over time and the W in millimeters, plotted in millions of years before present. B, Measure of the branching angle (BA) in degrees in branching and network forms over time, plotted in millions of years before present. The trend line is represented by the second-order polynomial best-fit line for the network forms. Black solid shapes and error bars for both plots represent the 95% confidence interval of the data set per time period, and the ages represent the median age of the estimated age range of the specimen. Light gray shapes represent the raw data. Geologic timescale provided for clarity.
The MI follows the same pattern seen in the meandering forms (Fig. 5B), with values starting low in the Paleozoic and increasing dramatically during the Late Cretaceous. The range of the branching forms is much more restricted, however, so it is not possible to confirm that the branching forms follow the same pattern during the Oligocene, because there are no Oligocene branching specimens. However, the branching specimens do help confirm the repressed MI values during the later Paleozoic that are seen within the meandering forms.
The DBox, W, and BA all show a very similar pattern across the analyzed samples (Fig. 6). For each of the analyses, the Paleozoic samples fall within the range of the Mesozoic–Cenozoic samples without any discernible trend visible across the time interval analyzed. Unfortunately, the small number of samples makes it difficult to identify more patterns with the data. One interesting pattern is that the BAs in the branching forms do not follow the same trend as the BAs in the network forms (Fig. 6B).
Network Forms
The study set of network forms consists of 182 specimens, ranging in age from the Fortunian (earliest Cambrian) to the modern, including the ichnogenera Gordia, Megagrapton, Paleodictyon, and Squamodictyon (Fig. 2). The measures of BA, network tortuosity (NT), W, MS, and DBox were plotted against the age of the specimens to determine any noticeable trends (Figs. 6B, 7, and 8).
Figure 7 Results of analyses of network forms. A, Measure of NT over time, plotted in millions of years before present. B, Measure of W in millimeters over time, plotted in millions of years before present. Black solid shapes and error bars for both plots represent the 95% confidence interval of the data set per time period, and the ages represent the median age of the estimated age range of the specimen. Light gray shapes represent the raw data. Geologic timescale provided for clarity. The darker gray background and arrows are included to help illustrate these changes in values.
Figure 8 Additional results of analyses of network forms. A, Measure of MS over time, plotted in millions of years before present. B, Measure of the fractal dimension calculated by the box method (DBox) over time, plotted in millions of years before present. Black solid shapes and error bars for both plots represent the 95% confidence interval of the data set per time period, and the ages represent the median age of the estimated age range of the specimen. Light gray shapes represent the raw data. Geologic timescale provided for clarity.
The plot of BA over geologic time (Fig. 6B) shows one of the strongest trends of any of the analyses performed. The trend line is represented by the second-order polynomial best-fit line. Possible BAs can range from ~5°, as a minimum measurable angle, up to 120°, as the maximum angle possible. A measure of 120° would represent a perfect juncture of three burrows, like a “Y” junction in which every angle is equal. While the branching forms had average BAs ranging from 40° to ~100° for their occurrence period, the network forms showed a distinct trend through time. As the network forms occur from the Cambrian through the Early Cretaceous, the average BA slowly increases from ~55° up to an average of 90°. As with the other forms, the range of values for each time slice decreases closer to the Cenozoic samples. Toward the Cenozoic, the maximum BA levels off at ~90°, even though a perfect hexagonal mesh would have an average BA of 120°.
The NT plot over geologic time (Fig. 7A) indicates that there was a slow decline in NT values up until the Late Cretaceous. The Paleozoic data again lack the uniformity of the Late Cretaceous and Cenozoic data; however, there is a slightly decreasing trend for the Paleozoic data. During the Late Cretaceous and Cenozoic, the decrease steepens, and the values become less scattered. An idealized NT for a perfect hexagonal mesh would be around 1.2.
The W follows a pattern similar to that of NT (Fig. 7B); however, the shift point is different. The Paleozoic and Mesozoic data decrease steadily until the Paleocene/Eocene boundary. Then, at the start of the Eocene, the values shift to a larger W, and both increase in size and stop following any sort of trend. The W has a wide range in values in the early Paleozoic that persist until most of the data range is reduced and remains fairly well constrained in the Silurian. This narrow range of values continues up until the Eocene, when the range of values increases fourfold.
The plot of MS over time (Fig. 8A) exhibits an obvious trend up through the Jurassic. Within the Cretaceous, the MS numbers plateau, but there is no change in evolutionary trend. Right at the Cretaceous/Paleocene boundary, however, there is a sharp shift in values. The average MS increases suddenly from ~5 to ~8. After the increase, the average MS stabilizes through the Paleocene and Eocene and then starts to decrease again up to the modern.
The final graph shows DBox through geologic time (Fig. 8B). This is similar to the MS plot, because it does not show any strong trends through time. From the Cambrian to the end of the Mesozoic, DBox remains mostly stable, based on the 95% confidence interval. There is a noticeable shift in DBox values just after the Jurassic/Cretaceous boundary, where the DBox values drop suddenly, then return to their previous values. Then, right before the end of the Cretaceous, the DBox values increase suddenly to an average of almost D=1.7. This increase is unusual, because the majority of DBox measurements across all graphoglyptids rarely went over D=1.7 (Lehane and Ekdale Reference Lehane and Ekdale2013a). Following the Cretaceous/Paleocene boundary, the DBox values take a sharp decline, averaging only 1.5. From there they average out or climb slightly.
Discussion
Previous Studies
The deep sea has been a focal point of discussions of evolutionary theory for many years (i.e., Dayton and Hessler Reference Dayton and Hessler1972; Gage and Tyler Reference Gage and Tyler1991; Emig and Geistdoerfer Reference Emig and Geistdoerfer2004). Sanders (Reference Sanders1968, Reference Sanders1969) suggested that the relative stability of environmental conditions in the deep-sea allowed benthic ecosystems to flourish and evolutionary transformations of benthic organisms to be practically nonexistent. Seilacher (Reference Seilacher1974, Reference Seilacher1977) expanded this hypothesis into behavioral evolution using trace fossils by invoking ichnospecies as a proxy for faunal diversity. Seilacher’s analyses showed that the number of ichnotaxa of flysch trace fossils expanded slowly over time until the Cretaceous, when they exploded in diversity (Fig. 9). The slow evolution in the Paleozoic and Mesozoic was attributed to the time–stability hypothesis of Sanders (Reference Sanders1968, Reference Sanders1969). However, it was noted that the graphoglyptid trace makers, and therefore the traces they produced, were evolving into smaller forms, perhaps to optimize their intake of the food resources in the sea floor. Seilacher (Reference Seilacher1974, Reference Seilacher1977) suggested that this evolutionary trend occurred until the Late Cretaceous, when the evolution of angiosperms provided abundant food material for the deep-sea endobenthos or increased amounts of calcareous ooze on the deep-sea floor caused a rapid behavioral diversification.
Figure 9 Previous analyses made by Seilacher (Reference Seilacher1974) and Uchman (Reference Uchman2003) to depict the evolution of deep-sea trace fossils based on the number of ichnospecies throughout the Phanerozoic. The Seilacher data include all flysch trace fossils (black circles), while the Uchman data are presented for both all flysch trace fossils (gray squares) and just graphoglyptid ichnospecies (black triangles).
Crimes and Fedonkin (Reference Crimes and Fedonkin1994) agreed that ichnogeneric diversity accelerated in the Cretaceous; however, they disagreed with the claim that there was any type of behavioral optimization taking place. By highlighting a few select trace fossil forms from various time periods, Crimes and Fedonkin (Reference Crimes and Fedonkin1994) had hoped to show that the increase in available knowledge in the literature proved the gradual evolutionary optimization hypothesis untenable. Although they presented an interesting hypothesis, there was a dearth of data; they presented only one or two trace fossil morphologies per time period.
Uchman (Reference Uchman2003, Reference Uchman2004) followed these earlier works by compiling the largest flysch trace fossil data set to date, analyzing more than 100 articles and localities to identify any significant trends involving ichnospecies diversification over time (Figs. 2 and 13 and Supplementary File 1). By considering trace fossil producer speciation rates, Uchman supported Seilacher’s hypothesis that diversity slowly increased throughout the Paleozoic and Mesozoic, although not as continuously as had been previously thought. Uchman surmised that the diversity of the graphoglyptids and other flysch trace fossils waxed and waned throughout the Paleozoic and Mesozoic as a result of various factors, including glaciation, deep-ocean warming, increased competition for food, and tectonic episodes, such as changes in mid-ocean rifting and subduction zones. He also noted the expansion of trace fossil diversity in the Late Cretaceous, especially during the Turonian stage (Uchman Reference Uchman2003, Reference Uchman2004). The initial diversification was again attributed to the appearance of angiosperms and/or increased calcareous oozes. According to this hypothesis, the expansion continued until the Eocene optimum because of the advent of oligotrophic conditions. Following the Eocene maximum was a decrease in the Oligocene, likely due to the Eocene/Oligocene boundary crisis, an interval of time with significantly decreased ocean bottom water temperatures. According to Uchman, improved conditions during the Miocene did not increase graphoglyptid diversity, so the graphoglyptid diversity remained constant. The niche optimization and miniaturization that was predicted for the graphoglyptids by Seilacher (Reference Seilacher1977) was questioned by Uchman (Reference Uchman2003, Reference Uchman2004).
Uchman’s (Reference Uchman2003, Reference Uchman2004) analyses focused on Paleodictyon and Squamodictyon as one uniform group, which is similar to how they are being examined here. The results showed evolutionary patterns that were discerned for different time ranges (Paleozoic, Paleogene, and Mesozoic/Miocene), resulting in different evolutionary trends for each time range. These trends were attributed to the evolution of the trace makers.
Implications from the Current Study
Basis of the Study
A closer look at the trace fossils was necessary to get a better understanding of how graphoglyptids have changed through time. Simply cataloging the number of ichnogenera per time interval is not sufficient to fully document how burrowing behavior has changed. The goal of this study is to examine the trace fossils based solely on their topological forms: meandering, spiraling, branching, and network forms. Each of these types of forms were unique enough that they required unique behaviors to create them. Therefore, if any behavioral evolutionary trends truly exist, they most likely would occur within a topological form rather than across different forms. This indicates that behavioral evolution is being affected by environmental shifts influencing the trace makers. Any environmental shifts would impact the trace makers in each topological group differently, indicating that each topological group is being created by distinct groups of organisms, which likely are not closely related.
The best way to determine any changes in behavior over time is to have a robust data set that is evenly distributed across time. This, however, is not possible in most geological and paleontological studies. Within this data set, the Paleozoic to mid-Mesozoic sample sets are often widely spaced and sparsely populated. This makes identification of any trends within the Paleozoic difficult, if not impossible. The dearth of data may be because of the limited number of turbidite deposits found in these times (Supplementary File 1), or it could reflect the actual paucity of the graphoglyptid trace fossils during this time interval. The Late Cretaceous to the Miocene data set, however, is extremely robust, and analytical results represent actual patterns in the fossil record. The Miocene samples are the latest known graphoglyptid trace fossils found in the rock record. Only modern examples are known after the late Miocene.
Diversification of Graphoglyptids
Of the several hypotheses proposed for the behavioral evolution of graphoglyptid trace fossils, diversification is the most widely accepted. Initially, diversification was identified simply by tabulating the number of ichnogenera per time period (i.e., Seilacher’s approach). This approach can be expanded to include diversification of morphologies within a topological category. Using Seilacher’s (Reference Seilacher1974) and Uchmans’s (Reference Uchman2004) previous diversity estimates (Fig. 9), there is a distinct decrease in the number of turbidite deposits containing graphoglyptids from the Carboniferous to the beginning of the Late Cretaceous. Seilacher reported no data for this time interval, while Uchman did include data from this interval, but the number is much less than before the Carboniferous and after the Late Cretaceous. The decrease in ichnodiversity means that diversification did not simply increase throughout the Phanerozoic. The decrease means that the diversity dropped dramatically during the second half of the Paleozoic for an unknown reason, but the lack of data makes the exact timing and cause of the diversity change difficult to pinpoint. What is noticeable is that the large range of morphological values (the 95% confidence interval) seen in the early Paleozoic (Figs. 4, 6B, 7, and 8) is less in the specimens found in the late Paleozoic and Mesozoic up until the Late Cretaceous (Figs. 5, 6, 7, and 8). Environmental pressures (e.g., changes in temperature, acidity, or chemical composition of the water and/or sediment) imposed on the graphoglyptid trace makers during this time also could account for the restricted diversity of forms. Changes in local environmental conditions are known to isolate and fragment deep-sea benthic communities (Rex and Etter Reference Rex and Etter2010). Slight changes in pressure, water circulation, and environmental fluctuations limit the extent to which species can proliferate and isolates the regions in which each benthic species is suitably adapted.
Starting in the Late Cretaceous and proceeding through the Miocene, there is a shift not only in the abundance of graphoglyptids but also in their morphological forms. Diversity appears to begin increasing exponentially at this point, as can be seen in the increased range of the MI in both meandering and branching forms (Fig. 5B), τ in meandering forms (Fig. 5A), and range of MS in network forms (Fig. 8). This observation supports the hypothesis that the graphoglyptid geometry was differentiating over time; however, it was not part of a linear increase in diversity from the Cambrian through later times.
Miniaturization of Graphoglyptids
Miniaturization of burrow sizes through time can be detected by measuring W for meandering and network forms (Figs. 4A and 7B). The meandering and network forms are the most robust of the four topological data sets, and within those two groups W was decreasing through the Paleozoic and Mesozoic. For the meandering forms, this decreasing trend continued through the Cenozoic as well and can be seen within the DBox trend (Fig. 4B). For the network forms, there is a strong shift at the Paleocene/Eocene boundary, where the average W shifts from a steadily decreasing trend with a sharp increase to larger values with greater range. Another indication of the size of the burrows in network forms is the MS. Average MS should be decreasing if the network forms are miniaturizing over time (Fig. 8A). During the Paleozoic, it is difficult to tell if in fact they are decreasing in size, but it is clear that during the Cretaceous, there is a stabilization of values at the low end of their range followed by an increase in the values and range across the Cretaceous and into the Paleocene. In essence, even though network forms may be decreasing in size over the most of the Paleozoic and Mesozoic, there is a clear increase in size during the Cenozoic (as also noted by Uchman Reference Uchman2003).
Optimization and Perfection of Graphoglyptids
Optimization is another major trend that has been suggested in the literature. Optimization has many different connotations that depend on the type of feeding strategy being discussed. For deposit feeding, optimal foraging strategy generally refers to the factors that govern how food-gathering activity is optimized (Mårell et al. Reference Mårell, Ball and Hofgaard2002; Sims et al. Reference Sims, Reynolds, Humphries, Southall, Wearmouth, Metcalfe and Twitchett2014). The optimal foraging strategy for deposit feeders entails obtaining the most nutrients out of the sediment with as little effort as possible (Charnov Reference Charnov1976; Schneider Reference Schneider1984). Over the time interval from the Cambrian through the Late Cretaceous, meandering and branching forms become increasingly intricate. This is seen in the increase in τ and MI (Fig. 5). More intricate patterns could imply a higher ratio of amount of food eaten to amount of sediment searched.
Optimal foraging strategy does not apply to agrichnial (farming or gardening) burrows, in which the occupant of the burrow allows bacteria or fungi to grow on the walls of the burrow and then feeds on that crop (Lehane and Ekdale Reference Lehane and Ekdale2013a,Reference Lehane and Ekdaleb). For these kinds of burrows, like Paleodictyon, what is an optimized form? There are distinct trends in the evolution of network graphoglyptids to suggest that the trace makers were standardizing the shapes they were creating. Unlike the meandering and branching traces, it is easier to identify an idealized morphology for the network traces. In network forms (Fig. 6B), there is a trend of increasing BAs starting in the Cambrian and continuing into the Late Cretaceous, at which point the BAs average around 90°. This trend indicates that the network forms were becoming more consistent in their morphology.
Environmental Factors
It is evident that major environmental changes were occurring in the oceans in the Late Cretaceous continuing into the Eocene (Figs. 4, 5, and 7). Previous studies of large-scale changes in graphoglyptid diversity pointed to the Late Cretaceous alone as the time at which that diversification started (Seilacher Reference Seilacher1977). Analyzing changes in graphoglyptid morphologies through time, as opposed to simply tabulating the number of ichnospecies, indicates that there were separate shifts in morphology in both the Late Cretaceous and the Eocene. The changes that took place in the Late Cretaceous may have resulted from either the evolution of the angiosperms or an increase of calcareous ooze deposition on the sea floor, as was hypothesized by Seilacher (Reference Seilacher1974) and Uchman (Reference Uchman2003). Results from this study offer no clear-cut determination between these two hypotheses, nor do they indicate whether the cause was something else entirely.
A possible explanation for the explosion of graphoglyptid morphological diversity during the early Eocene, which produced the most abundant graphoglyptid occurrences in geologic history (Uchman Reference Uchman2003), was the Paleocene–Eocene Thermal Maximum (PETM), an event of abrupt, high-magnitude, global warming. The PETM climatic effects influenced even the deep sea, where an increase of bottom-water temperature of 4–5°C (Higgins and Schrag Reference Higgins and Schrag2006) occurred along with oxygen depletion (Rex and Etter Reference Rex and Etter2010) and water-column stratification (Bralower Reference Bralower2002). These changes had dramatic impact on the deep-sea benthos. The temperature increase and oxygen depletion at the PETM were identified as probable causes of a 30–50% extinction of benthic foraminifera (Alegret et al. Reference Alegret, Ortiz and Molina2009; Rex and Etter Reference Rex and Etter2010). While the effects of the PETM apparently did not affect all organisms as dramatically as it did the benthic foraminifera, there was a significant loss of benthic life, as seen in the trace fossil record across the Paleocene/Eocene boundary (Cummings and Hodgson Reference Cummings and Hodgson2011; Rodríguez-Tovar et al. Reference Rodríguez-Tovar, Uchman, Alegret and Molina2011). The very cold (~4°C) waters in the deep sea facilitated the distribution of the benthic fauna by slowing their metabolic demands and allowing them to disperse over long distances. The warmer temperatures during the PETM limited this dispersal ability. The effects of the PETM led to a fragmentation of deep-sea ecosystems, where previously there was a more uniform environment. This fragmentation produced by shifting environmental factors spurred differentiation and speciation of the deep-sea benthos (Rex and Etter Reference Rex and Etter2010). The severe reduction of trace fossils at the PETM slowly recovered into the Eocene as nutrient availability improved and oxygen replenished, thus bringing a return of benthic life (Rodríguez-Tovar et al. Reference Rodríguez-Tovar, Uchman, Alegret and Molina2011). The warmer waters likely helped with the introduction of additional organic material, with biodegradation increasing as water temperatures increased (Rex and Etter Reference Rex and Etter2010). At this time of slow speciation and fragmented environment, it is likely that the behaviors of the deep-sea benthos also evolved to compensate for the environmental conditions.
Following the ichnodiversity maximum in the Eocene, there was a crash in diversity that may be attributed to the Eocene/Oligocene boundary crisis, which was associated with a drop in oceanic water temperatures (Uchman Reference Uchman2003). Other studies, however, suggest that even though this was the first major continental-scale ice accumulation of the Cenozoic, there was no appreciable drop in deep-sea temperatures (Lear et al. Reference Lear, Elderfield and Wilson2000). The growth of the Antarctic ice sheets caused a drop in global sea level, increased limestone erosion worldwide, increased deposition of inorganic carbon in the oceans, and deepened the carbonate compensation depth. The influx of large amounts of CaCO3 increased the alkalinity of the oceans and contributed to the deacidification, rapidly changing the oceanic chemistry (Merico et al. Reference Merico, Tyrrell and Wilson2008). Some studies show that chemical conditions affect the migration patterns of benthic marine invertebrate larvae, since the larvae search for specific chemical cues in determining their settlement sites (Pawlik Reference Pawlik1992). It is possible that a sudden change in the deep-sea water chemistry could have caused a reduction in the preserved graphoglyptid morphology diversity by forcing the trace makers into deeper waters not suitable for graphoglyptid preservation. As the oceanic environment became more uniform in the Miocene, graphoglyptid diversity once again increased.
Uniquely Evolving Topological Forms
It appears that the different topological forms of graphoglyptids evolved separately, even though several of the topological forms share similar morphological aspects. For example, the BAs in the branching forms and the network forms follow a distinctly different evolutionary pattern. The same is true for the τ of the meandering burrows and branching burrows, as compared to the NT of the network forms. This finding indicates that while the meandering burrows were becoming more complex, the network burrows were becoming less complex and more regular. The varying evolutionary reactions to different environmental stresses by each of the different burrow topologies suggests that there probably were different taxa of trace makers producing the different topologies.
Conclusion
Behavioral evolution is a topic that is not often studied because of the difficulties in finding tangible evidence of fossil behavior that is easy to interpret. Trace fossils provide the main avenue for studying behaviors of ancient organisms, and by studying the trace fossil record through geological time it is possible to determine how at least some types of behaviors have evolved. Graphoglyptids are a group of trace fossils ideal for studying behavioral evolution because of their typical occurrence in deep-marine settings along with their characteristic mode of formation as open-burrow systems, often preserved on the base of turbidite beds. Graphoglyptid trace fossil evolution has been studied by a few ichnologists, principally by tabulating the number of ichnogenera as a proxy for diversification of specific behavior. However, this type of strictly taxonomic approach does not take into account the morphometric variations displayed by different ichnogenera, which provide the focal point for the current study.
To use the entire array of graphoglyptid morphologies, this study analyzed graphoglyptid trace fossils within different topological form categories: meandering, spiraling, branching, and network. Each specimen, approximately 400 in total, was analyzed by several quantitative analytical techniques, including fractal analysis, BA, W, NT, τ, MI, and MS. Previous analyses concluded that graphoglyptids were evolving for purposes of optimization of feeding patterns, while getting smaller through time until the Late Cretaceous, when a sudden increase in diversification occurred. This interval of rapid diversification of graphoglyptid ichnotaxa was attributed by previous workers to either the evolution of the angiosperms on land or the sudden increase in foraminiferal ooze in the deep sea.
Results of this analytical study suggest that understanding the behavioral evolution of the graphoglyptid trace makers is more complicated than simply documenting trends in ichnotaxonomic diversity through time. The different topological forms evolved separately from each other, and while some were following previously proposed evolutionary patterns, others were not. In general, the pattern of graphoglyptid behavioral evolution can be divided into three different time intervals: (1) Paleozoic and Mesozoic before the Late Cretaceous, (2) Late Cretaceous to the end of the Eocene, and (3) post-Eocene. During the interval before the Late Cretaceous, graphoglyptids were miniaturizing and decreasing in diversity. Starting in the Late Cretaceous, the graphoglyptid diversity was greatly expanding, while meandering forms continued to miniaturize, and network forms enlarged. During this interval, meandering forms also were increasing in geometric complexity, while network forms were becoming more regular in geometry.
Evolutionary pressures in the deep-marine environment may have included angiosperm evolution or an increase in foraminiferal ooze during the Late Cretaceous, as suggested by previous workers, or it may have included increased bottom-water temperatures during the PETM and ocean chemistry changes during the Eocene/Oligocene boundary event, as this current study suggests. Whatever the case, it is important to note that environmental changes of one sort or another affected the different topological groups of graphoglyptids differently. This observation suggests that graphoglyptid burrows belonging to each of the topologic categories may have been created by different taxa of trace makers, and this finding casts doubt on the hypothesis that all graphoglyptid trace fossils were created by closely related animal species.
Acknowledgments
The authors thank R. Irmis, L.Tapanila, A. Uchman, W. Miller III, and an unnamed reviewer for useful comments on the manuscript; A. Wetzel, T. Good, P. Dentzien-Dias, A. Uchman, J. Gruza, E. Gierlowski-Kordesch, M. Goodwin, W. Miller III, M. Gingras, W. Obcowski, and L. Buatois for help in field and museum work and sample identification; R. Bruhn, L. van Vliet, K. van der Voort Maarschalk, A. Baucon, J. de Gibert, and G. Mackie for mathematical and computer programming assistance; and M. Bednarz and S. Evers for help in translations. Funding for the research was provided in part by an ExxonMobil Geoscience Grant and the Graduate Research Fellowship provided by the University of Utah (to J.R.L.).
Supplementary Material
Supplemental material deposited at Dryad: doi: 10.5061/dryad.1tf47
