Optic Foramen Morphology and Activity Pattern in Birds

Study Animals

The optic foramina were observed for 177 avian families across 27 different orders (see Table 1). Of these, 17 families exhibit Type 1 optic foramina exclusively (see later), including Accipitridae (hawks), Aegothelidae (owlet-nightjars), Apterygidae (kiwis), Balaenicipitidae (shoebills), Cacatuidae (cockatoos), Caprimulgidae (nightjars), Cathartidae (New World vultures), Falconidae (falcons), Nyctibiidae (potoos), Podargidae (frogmouths), Psittacidae (parrots), Rheidae (rheas), Steatornithidae (oil birds), Strigidae (true owls), Struthionidae (ostriches), Tytonidae (barn owls), and Upupidae (hoopoes). Specimens observed and measured for this study were obtained from the Divisions of Birds at the National Museum of Natural History (Smithsonian Institution, Washington, DC) and the Field Museum of Natural History (Chicago, IL), and the Department of Ornithology at the American Museum of Natural History (New York, NY). Only adult specimens were included in our analyses.

Of the 17 families exhibiting a Type 1 optic foramen (see definition below), measurements were taken for 15 families. For the two families that were not measured, Balaenicipitidae and Apterygidae, the optic foramina are clearly Type 1, but they are present on the midline of orbits so deep that they are not accessible to calipers. A total of 775 specimens of 313 species of nocturnal and diurnal birds exhibiting a Type 1 optic foramen were measured (see Table 2).

Optic Foramen Coding

Birds were coded into one of four optic foramen types as follows (see Fig. 1): Type 1, a discrete optic foramen is present in each orbit that probably reflects the size of the optic nerve; Type 2, a single, central optic foramen is present that probably reflects the position of the optic chiasm (M.I.H., personal observation); Type 3, an optic foramen is present that, although discrete, is too large to reflect the optic nerve; and Type 4, no discrete optic foramen is present because the posterior aspect of the orbit is not ossified.

Activity Pattern Coding

All birds exhibiting a Type 1 optic foramen were coded into one of two different activity patterns from the literature (del Hoyo et al,1999): (1) diurnal, defined as active during the day in a photopic environment and (2) nocturnal, active at night in a scotopic light environment. No exclusively cathemeral (equally likely to be active at any time of day) or crepuscular (active during dawn and dusk) birds were included in this study; several owls that are active during dawn and dusk are primarily nocturnal and were therefore included in the nocturnal group. We define activity pattern as that time of the day when animals conduct the usual activities of life, and do not include occasional activities that take place at other times of day.

Measurements

Measurements were performed with digital calipers to the nearest 0.01 mm as follows, although not every measurement was available for every specimen (see Table 2 and Fig. 2). The maximum diameter of the optic foramina was measured. In some taxa, the optic foramen was not fully ossified around the perimeter in that the optic foramen was only partially enclosed. For these cases, the maximum diameter that was ossified on both sides was measured. In addition to optic foramen maximum diameter, three additional measurements were taken of each specimen where possible: (1) head length, (2) orbit diameter, and (3) the inner diameter of the sclerotic ring. Head length was measured to investigate the allometry of the optic foramen across birds, and was measured from the centerpoint of the junction of the keratinous sheath of the beak proper and the bony skull to the posterior-most point on the skull. Previous work has shown that both orbit diameter (Kay and Kirk,2000; Heesy and Ross,2001,2004; Kirk and Kay,2004; Kirk,2006a; Hall,2008b) and the inner diameter of the sclerotic ring (Hall,2008b; Schmitz,2009) differ between nocturnal and diurnal species. Orbit diameter is a bony correlate of the transverse diameter of the eye, a useful indicator of overall eye size, and nocturnal birds exhibit a larger eye (Ritland,1982; Hall and Ross,2007) and orbit diameter than diurnal birds (Hall,2008b). Nocturnal birds also consistently exhibit a larger inner diameter of the sclerotic ring, a close bony correlate of corneal size, probably as an adaptation to allow more light to enter the eye (Hall,2008b). Most birds do not have an ossified inferior orbital margin, in which case orbit diameter was measured from the center of the quadratojugal, closest to the beak margin, to that point on the superior orbital margin directly opposite (after the mammalian Orbitale inferius and Orbitale superius, as defined in Cartmill,1970). In specimens with an ossified inferior orbital margin (e.g., parrots), orbital diameter was measured from that point on the inferior orbital margin closest to the beak margin and the center of the quadratojugal bar to that point on the superior orbital margin directly opposite. The maximum inner diameter of the sclerotic ring was also measured.

Statistical Analysis

To determine if optic foramen size differs between nocturnal and diurnal birds, we used both comparisons between means and regression analyses. All of the data were log₁₀ transformed and tested for normality using the Kolmogorov-Smirnoff test with Lilliefors transformation to determine if parametric or nonparametric statistics were appropriate. We generated a box plot to demonstrate the differences of the optic foramen maximum diameter between nocturnal and diurnal birds (see results, Fig. 3). We also generated scatter plots to allow for further visual inspection of the data (see Results, Figs. 4-6). Because measurement error and natural variation affect both dependent and independent variables, reduced major axis (RMA) was the Model II line-fitting technique used in this study. RMA regression analyses do not assume that variance in either variable is more significant than the other, or that one is influencing the other (Ricker,1984; Rayner,1985; Sokal and Rohlf,1995; Warton et al.,2006). We also report ordinary least squares regression results to better allow comparisons of our results with other studies (e.g., Kay and Kirk,2000; Kirk and Kay,2004). RMA regressions were calculated in (S)MATR (Falster et al.,2006); all other analyses were computed in SPSS 15.0.

Phylogenetic Comparative Methods

Interspecific scaling relationships are significantly affected by phylogenetic history. Failure to account for this phylogenetic effect can lead to an increased risk of Type I errors (Harvey and Pagel,1991). We therefore adopted phylogenetically based comparative methods to incorporate phylogenetic information in our analyses.

Currently, there is no consensus concerning the phylogenetic relationships among avian orders and families. We therefore generated four different phylogenetic trees based on the interordinal relationships provided in Sibley and Ahlquist (1990), Cracraft et al. (2004), Livezey and Zusi (2007), and Hackett et al. (2008). Because these studies focused on interordinal and -familial relationships, resolution at the species level was provided by additional sources (Brown and Toft,1999; Riesing et al.,2003; Griffiths et al.,2004; Wink and Sauer-Gurth,2004; Lerner and Mindell,2005; Barrowclough et al.,2006; Wink et al.,2008; Wright et al.,2008). Genera and species not present in these studies were omitted from our phylogenetic tree because incorrect placement of species can result in an increased risk of error in calculating correlation coefficients (Symonds,2002). It should also be noted that despite the resolution provided by the additional studies cited earlier, several polytomies were present across the tree.

Two types of comparative analysis were performed. First, we used Felsenstein's (1985) independent contrasts approach to derive phylogeny-corrected estimates of the slope of the scaling relationships as well as the amount of variation explained (r²). The reconstructed phylogenies and the data set were entered into Mesquite (v. 2.6), a modular comparative analysis program (Maddison and Maddison,2009). As with the analysis of species as independent data points (see earlier), all of the data were log-transformed before analysis. The independent contrasts were then calculated using the PDAP:PDTREE module of Mesquite (Midford et al.,2005). Because the phylogeny was derived from multiple sources that used different characters to derive their trees (e.g., mitochondrial DNA, morphology, DNA hybridization), branch lengths were given arbitrary values according to several different models. We tested each of these arbitrary branch length models for their ability to adequately standardize the data following Garland et al. (1992). Nee's (Nee, cited in Purvis,1995) arbitrary branch length model adequately standardized all of the data for the Sibley and Ahlquist (1990) and Livezey and Zusi (2007) phylogenies. This model was also appropriate for orbit diameter and sclerotic ring inner diameter in the Hackett et al. (2008) phylogeny and head length in the Cracraft et al. (2004) phylogeny. The only arbitrary model that successfully standardized the remaining comparisons was Grafen's (1989) rho model (with ρ = 0.5). Once the contrasts were standardized, least-squares and RMA regression lines were forced through the origin (Garland et al.,1992) as implemented in the PDAP:PDTREE module, and the slopes and correlation coefficients of these lines calculated.

The second type of comparative analysis adopted was the phylogenetic generalized least-squares (PGLS) approach as implemented in the MATLAB program Regressionv2.m (available from Garland et al.,1992). The data were log-transformed and we also included activity pattern as a categorical variable and analyzed the data as a mixed regression model. We applied two models of evolutionary change as implemented in Regressionv2.m: Brownian motion (PGLS) and Ornstein-Uhlenbeck (Lavin et al.,2008; Swanson and Garland, 2009). Partial F tests and Akaike Information Criterion were then used to determine which model best fit a multiple regression consisting of optic foramen diameter as the dependent variable and one of the scaling variables (i.e., head length, orbit diameter, or sclerotic ring inner diameter) and activity pattern (nocturnal or diurnal) as independent variables. The results of these analyses will determine whether there are significant differences in relative optic foramen diameter between nocturnal and diurnal birds.

Comparisons Between Means

Log₁₀ optic foramen maximum diameter was not log-normally distributed (as indicated by a significant Kolmogorov-Smirnoff's test with Lilliefors' transformation, P = 0.000). Therefore, these data do not meet the assumptions of ANOVA, and a Kruskal-Wallis nonparametric alternative to ANOVA was performed to test for differences between the means of the two groups, yielding a significant difference (df = 1, P = 0.000, chi-square = 110.626).

Inner Diameter of the Sclerotic Ring

Optic foramen maximum diameter scaled against the inner diameter of the sclerotic ring with negative allometry (slope < 1) across all species for all OLS regressions, but with isometry (slope approximates 1) for all RMA regressions (Table 3). When the species were separated according to activity pattern, regression analyses revealed significant differences between nocturnal and diurnal species. These activity pattern specific regression models tended to explain more variation (i.e., higher r²s) than examining variation across all species, and Fig. 4 shows that the two groups are clearly shifted along the y-axis. In terms of the specific scaling relationships themselves, diurnal species exhibited a negative allometric relationship when species were used as independent data points. The scaling relationship did, however, vary from negative allometry to isometry in our independent contrasts analyses, depending upon whether OLS (negative allometry) or RMA (isometry) models were used (Table 3). Similarly, the scaling relationship in nocturnal species varied from negative to positive allometry depending upon whether phylogenetic information was incorporated or not and the regression model used (Table 3).

Because the slopes of the regression lines for nocturnal and diurnal species are not identical, ANCOVA analyses were not appropriate to determine if the differences in elevation were statistically significant (Sokal and Rohlf,1995). Therefore, in our analysis of species as independent data points, a single RMA line was calculated for all groups (y = 1.14 × −0.701), and residuals for all data points were calculated from this line for the y-variable, log₁₀ optic foramen maximum diameter. These residuals were not log-normally distributed, so a Kruskal-Wallis nonparametric alternative to ANOVA was performed to test for differences between the means of the three groups, yielding a significant result (asymp. sig. = 0.000, df = 2, chi-square = 24.363). When we constructed a mixed linear model using activity pattern and sclerotic ring inner diameter as covariates of optic foramen maximum diameter, we also recovered a significant difference in activity pattern (Table 4). This was further corroborated by the PGLS approach using two models of evolutionary change across all four phylogenetic trees (Table 4). Thus, nocturnal species have significantly smaller optic foramina than diurnal species, relative to the sclerotic ring inner diameter.

Orbit Diameter

Regression analyses of optic foramen maximum diameter against orbit diameter across all species yielded a negative allometric or isometric relationship, depending upon whether an OLS or RMA model was used or whether phylogenetic information was included or not (Table 3). As with the analysis of sclerotic ring inner diameter, optic foramen size relative to orbit diameter differs between diurnal and nocturnal species (Fig. 5). Separate regression analyses of the two activity patterns yielded similar patterns to our analyses of sclerotic ring inner diameter. Specifically, the relationship between optic foramen diameter and orbit diameter in diurnal species scaled with negative allometry across all analyses, whereas the same relationship in nocturnal species varied from negative to positive allometry, depending on the model used (Table 3). As with the inner diameter of the sclerotic ring, the slopes for nocturnal and diurnal birds with orbit diameter on the x-axis were not identical in RMA space and therefore not appropriate for ANCOVA analyses. Therefore, we again calculated residuals from the RMA line, as described earlier. A Kruskal-Wallis test performed on the residuals also showed that nocturnal and diurnal birds occupy statistically different portions of the graph (asymp. sig. = 0.000, df = 1, chi-square = 132.356).

The mixed regression model indicated that relative to orbit diameter, nocturnal species have significantly smaller optic foramen diameters than diurnal species (Table 4). This significant difference was detected in all nine models and corroborates our analyses of sclerotic ring inner diameter, although orbit diameter explained slightly more interspecific variation (r² = 0.088–0.483 compared to 0.062–0.357).

Head Length

Regression analyses of optic foramen diameter against head length yielded broadly similar results. Just as with the previous analyses, in combined analyses of all species, the allometric relationship between optic foramen diameter and head length varied from negative allometry to approximately isometry, depending on the regression model used (Table 3). A plot of optic foramen diameter against head length (Fig. 6) also indicated some separation of nocturnal and diurnal species. Regression analyses within the two activity patterns yielded the same qualitative results as the previous analyses; optic foramen diameter scales with negative allometry to isometry within diurnal species and with negative to positive allometry within nocturnal species, depending on which regression model is used (Table 3). Again, nocturnal and diurnal birds exhibited heterogeneous slopes in RMA space, so we calculated residuals from a single RMA line to determine whether the activity patterns occupy statistically different portions of the graph. A Kruskal-Wallis analysis of the residuals yielded a positive result, as with the other variables (asymp. sig. = 0.000, df = 1, chi-square = 39.849). However, the separation between activity patterns, although statistically significant, is not as clear as with the other variables (Fig. 6). For example, there is so much overlap between the nocturnal and diurnal ranges that if one were to plot a single point, such as a fossil, it would be very difficult to determine activity pattern with confidence.

The mixed regression models also corroborated our previous findings. Regardless of whether phylogenetic information was incorporated or not, a significant difference in optic foramen size was detected between nocturnal and diurnal species; relative to head length, nocturnal species have a significantly smaller optic foramen than diurnal species (Table 4). Therefore, our conclusion is that relative optic foramen diameter is significantly smaller in nocturnal species than diurnal species, and this effect is robust to different regression and evolutionary change models.

Optic Foramen Types

For birds with an ossified posterior orbital wall, the three different optic foramen types probably all reflect the size of a portion of the visual pathway from the retina to the rest of the brain, either the optic chiasm or the optic nerve itself. For example, a single, central optic foramen (Type 2, Fig. 1B) probably approximates the size and position of the optic chiasm (M.I.H., personal observation). The Type 3 foramen (Fig. 1C) is likely a variant of the Type 2 optic foramen with the addition of a central strut that divides a single, central optic foramen into two, but is still the position of the optic chiasm. Indeed, many bird genera exhibit both Type 2 and Type 3 foramina, and occasionally some individuals of the same species exhibit a Type 2 and others a Type 3, indicating the possibility that the central strut observed in Type 3 foramina may sometimes be lost during museum preparation. There are obvious differences among species, genera, and families in the degree of ossification of the orbit, but why these variations occur in birds is currently unknown.

Type 1 Optic Foramen: Activity Pattern Analyses

Both regression statistics and comparisons between means show that the maximum diameter of the bird Type 1 optic foramen discriminates between nocturnal and diurnal birds. The optic foramen discriminates between nocturnal and diurnal primates (Kay and Kirk,2000; Kirk and Kay,2004; Kirk,2006a), and, as discussed previously, likely arises from differences in retinal summation between nocturnal and diurnal species. Because the optic nerve is the sole output of the retinal ganglion cell layer to the rest of the brain, the optic nerve of a diurnal bird was predicted to be larger because of a greater number of ganglion cells, an important adaptation for visual acuity. The opposite was predicted for nocturnal birds; nocturnal species were predicted to have fewer ganglion cells with a higher amount of retinal summation to improve visual sensitivity. This study supports both of these predictions, with the majority of diurnal birds exhibiting an absolutely larger Type 1 optic foramen. However, even though optic foramen size is statistically different between nocturnal and diurnal birds, it would still be very difficult to interpret a fossil for which optic foramen size was the only feature available, unless it plotted in the extremes (Fig. 3).

Regressions of the optic foramen maximum diameter against the inner diameter of the sclerotic ring, a close anatomical correlate of corneal diameter, also discriminates between diurnal and nocturnal birds (see Fig. 4). Previous work has shown that corneal diameter (Ritland,1982; Hall and Ross,2007), lens diameter (Schmitz,2009), and the inner diameter of the sclerotic ring discriminate between nocturnal and diurnal birds (Hall,2008b; Schmitz,2008,2009) and lizards (Hall,2008a,2009). Nocturnal birds probably possess a larger inner diameter of the sclerotic ring and corneal diameter to allow more light to enter the eye (e.g., Martin,1990). This study demonstrates that at a given corneal diameter, a nocturnal animal consistently demonstrates a smaller optic foramen, indicating greater retinal summation.

Our study also shows that optic foramen maximum diameter relative to orbital diameter, a bony correlate of the transverse diameter of the eye and a useful measurement of overall eye size, discriminates between nocturnal and diurnal birds (Table 3, Fig. 5). This robust difference was consistent across all of our analyses, despite that fact that previous work indicated that orbit diameter, either alone or in combination with the sclerotic ring, is not informative for activity pattern in birds (Hall,2008b). Orbit diameter is a useful variable in activity pattern analyses in primates (Kay and Cartmill,1977; Kay and Kirk,2000; Heesy and Ross,2001,2004; Kirk and Kay,2004) and clearly, when combined with optic foramen data, can be used to discriminate activity pattern in birds. That is, diurnal species have a significantly larger optic foramen relative to orbital diameter than nocturnal species. In other words, because orbit diameter is a proxy of eye size, when birds of a similar eye size are compared, the diurnal bird will consistently display a larger optic foramen. This observation supports the idea that diurnal birds exhibit less retinal summation, assuming that nocturnal and diurnal birds of a similar eye size possess similar total numbers of photoreceptors.

Interestingly, the variance explained by the variables investigated in this study is very different for diurnal and nocturnal birds; all r² values for allometric relationships specific to nocturnal species vary between 0.195 and 0.514, whereas those for diurnal species are between 0.547 and 0.717 (Table 3). Because the size of the optic nerve is thought to reflect the degree of retinal summation in an animal, the low nocturnal r² values may reflect different retinal summation strategies among nocturnal birds. For example, previous work has shown that some nocturnal owls exhibit absolutely larger axial lengths of the eye than even many diurnal birds (Ritland,1982; Howland et al.,2004; Hall and Ross,2007), suggesting that they are attempting to maintain the maximum visual acuity possible in the context of increased visual sensitivity. In contrast, many nightjars (Caprimulgidae) have shorter axial eye lengths relative to corneal diameter, which suggests that the nightjar strategy is to increase sensitivity without maintaining much acuity (Ritland,1982; Hall,2005; Hall and Ross,2007). Perhaps high acuity is not necessary for trawling for insects in nightjars (Martin,1990; del Hoyo et al.,1999), whereas owls that hunt on the wing under scotopic conditions (Martin,1990; del Hoyo et al.,1999) require both sensitivity and moderate acuity to locate individual prey items. The wide distribution of optic foramen diameters may reflect these different strategies. Diurnal birds, however, always function under photopic conditions and are not forced to make compromises between sensitivity and acuity. Therefore, diurnal birds are all free to limit retinal summation and increase ganglion cell input to the brain to maximize visual acuity, and hence have a more consistently large optic nerve and a limited distribution of optic foramen sizes.

What About Owls?

The robust separation between nocturnal and diurnal slopes provided by optic foramen diameter does not extend to owls. In this sample, diurnal owls (7 spp.) are all found within the same distribution as nocturnal owls. However, several other groups of nocturnal birds are part of this analysis in addition to the nocturnal owls, including the nightjars, owlet-nightjars, potoos, frogmouths, the oilbird (Steatornis caripensis), and the enigmatic kakapo (Strigops habroptilus), all of which plot as predicted within the nocturnal range. Besides the diurnal owls, there are no diurnal birds that plot within the nocturnal range. So, why do diurnal owls display a different pattern?

Diurnal owls also often follow a nocturnal eye shape pattern, with a larger corneal diameter relative to a shorter axial length of the eye; again, virtually no other diurnal birds follow this pattern (Hall,2005; Hall and Ross,2007). Although the evolutionary history of owls is not well understood, an examination of activity pattern across the owl clade suggests that extant diurnal owls probably descend from a nocturnal ancestor (Wink et al.,2008; Gutiérrez-Ibáñez et al., in preparation). It may be that once owls evolved a nocturnal eye shape with a relatively large corneal diameter, the subsequent evolution of diurnality did not provide selection pressure to change eye shape to a relatively smaller corneal diameter. As Martin (1982,1990) has argued, perhaps it is not correct to discuss “nocturnal” and “diurnal” eye shapes, but rather “arrhythmic” and “diurnal” eye shapes; regardless of eye shape, in photopic conditions an animal potentially can reduce pupil size without affecting overall eye shape to limit the amount of light entering the eye. However, this argument does not extend to optic foramen diameter. The optic nerve cannot be larger than the optic foramen. The small size of the diurnal owl optic foramen, and therefore optic nerve, strongly suggests that these birds retain a nocturnal retinal summation strategy, with retinal wiring for maximized visual sensitivity, not visual acuity. This is surprising in the case of owls such as the snowy owl (Bubo scandiacus), a totally diurnal owl that routinely competes with hawks and falcons for small mammal prey under photopic conditions (Parmalee, 1992; del Hoyo et al.,1999).