Factors underlying the Allee effect are still heavily debated in ecology. For plants that rely on pollinators for seed production, decreases in conspecific aggregation may reduce attractiveness to floral visitors and lead to an Allee effect. However, floral visitors often differ in their pollination effectiveness; hence, the likelihood of an Allee effect in plant fecundity may depend on how various flower visitors respond to plant aggregation.
We tested for Allee effects on fecundity of individuals across two years in the self-incompatible perennial, Kniphofia linearifolia Baker (Xanthorrhoeaceae), which has two distinct types of pollinator, birds and native bees.
For this, we used three measures of aggregation; population size, density and isolation. We made replicated pollinator observations in populations of various aggregations and quantified fecundity in these populations. To determine the differences in pollinator effectiveness and assess their contribution to fecundity, we selectively excluded bird visitors from K. linearifolia in these populations.
We found that population size, but not density or isolation distance, was associated with increased bird abundance and seed set in one of the two years of the study. Bird visitation rate increased with increased plant aggregation within populations. Fruit set and seed set per flower were positively related to bird visitation rate. The difference in seed set per flower between bird-excluded and unmanipulated plants increased with increasing population size. Although birds were much less frequent visitors than bees (on average 2.1 visits plant−1 h−1 compared to 57.5 visits plant-1 h−1), selective exclusion experiments indicated that birds are consistently the more effective pollinators of this species, and therefore most likely to influence fecundity.
Synthesis. In this system, characterised by an Allee effect on plant fecundity, birds were the most effective pollinators, responded positively to plant aggregation and were associated with increased fecundity. Therefore, the responses of effective pollinators to plant aggregation may be a factor that underlies Allee effects on plant fecundity.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The positive effect of increased aggregation of conspecifics on individual fitness, the Allee effect (Allee 1938), is known to be important in governing the per capita fecundity and overall growth of plant populations (Lamont, Klinkhamer & Witkowski 1993; Groom 1998; Hackney & McGraw 2001; Knight 2003; Brys et al. 2004). Per capita fecundity is an important component of population demography as it is often positively associated with the long-term viability of populations (Stephens, Sutherland & Freckleton 1999; Courchamp, Berec & Gascoigne 2008). The Allee effect has often been defined, particularly by zoologists who seldom have access to individual fecundity data, as a decrease in population growth rate with decreasing population density (Courchamp et al. 2008). However, Stephens, Sutherland & Freckleton (1999) emphasised that components of overall fitness, such as per capita fecundity, can relate to population survival and growth rate, and that one should therefore distinguish between ‘component’ Allee effects and ‘demographic’ Allee effects. Indeed, studies of component Allee effects are essential for identifying the underlying basis of demographic Allee effects. Recently, Feldman & Morris (2011) demonstrated that increases in fecundity at high densities (a component Allee effect) might be obviated by increased survival rates at low densities, thus neutralising the demographic Allee effect. However, their study did not examine the effect of population size on survival rates, which may be decoupled from the effect of population density (Groom 1998; Leimu et al. 2006). Indeed, Allee (1938) used the more general term ‘aggregation’ to describe how individuals can suffer from ‘under-crowding’ in populations of low size and/or density. Therefore, several recent studies have attempted to control for both density and size when investigating Allee effects, and found either a density or population size effect or both (Groom 1998; Ward & Johnson 2005; Johnson, Torninger & Ågren 2009). In this paper, we present data that show that a component Allee effect on fecundity is due to increasing population size, and not density or isolation in an animal-pollinated plant, and that this effect may depend on the type of pollinator available.
Due to increasing anthropogenic influence on natural habitats, it is becoming increasingly important to quantify and understand the underlying mechanisms that drive Allee effects in natural populations (Aizen, Ashworth & Galetto 2002). This is because there may be increased extinction risk for both plants and animals below a certain level of conspecific aggregation (Groom 1998; Brashares, Werner & Sinclair 2010), and extinctions of plants and pollinators may occur in tandem (Biesmeijer et al. 2006). Generally, Allee effects in the reproduction of animal-pollinated plants can occur by: (i) increased pollinator attraction with increases in the aggregation of conspecific individuals (Sih & Baltus 1987), (ii) increased likelihood that conspecific pollen is transferred to conspecifics (Silander 1978; Groom 1998) and (iii) increased uniparental and biparental inbreeding with decreasing aggregation (Aizen & Feinsinger 1994; Ågren 1996; Johnson et al. 2009), or a combination of these factors. Consequently, Allee effects can arise due to insufficient pollination quality and quantity in smaller populations, which can negatively affect fecundity. While many studies have quantified increased fecundity with increasing population size and density (Brys et al. 2004; Spigler & Chang 2008), or decreasing isolation (Steffan-Dewenter & Tscharntke 1999), no study has addressed the responses of different pollinators to the same mosaic of aggregation of a single species, and their effects on plant fecundity. Pollinators often differ in their effectiveness at inducing seed set (Ne'eman et al. 2010) and there is frequently high stochasticity in the relative abundance of different pollinators among plant populations (Wilson & Thomson 1991), which may explain differences in fecundity (Schemske & Horvitz 1984; Cosacov, Nattero & Cocucci 2008).
In this study, we investigate the reproductive biology of the perennial, self-incompatible geophyte Kniphofia linearifolia (Baker) (Xanthorrhoeaceae), which has two distinct groups of pollinator; birds and native bees (Brown, Downs & Johnson 2011). Both pollinator groups differ in their morphology and foraging behaviour, with birds foraging for nectar and capable of flying longer distances both within and between populations. Bees tend to fly shorter distances mainly within populations, foraging predominately for the exposed pollen, and occasionally for nectar. We explored the effects of three commonly used population-scale measures of aggregation: population size, density and isolation on pollinator visitation and plant fecundity. Here, we test whether the likelihood of an Allee effect on plant fecundity is dependent on the type of pollinator available in a population. Specifically, we address the following questions: (i) Is the rate of bird and bee visitation to K. linearifolia plants affected by the size, density and isolation of flowering populations? (ii) Does plant fecundity vary according to population size, density, isolation and the rate of pollinator visitation? (iii) Does fecundity vary between open-pollinated and bird-excluded plants according to the size, density and isolation of populations?
Materials and methods
Study Species and Populations
Kniphofia linearifolia (common name: red-hot poker) is a widespread member of the Xanthorrhoeaceae (now incorporating the former Asphodelaceae) in South Africa. It produces long, narrow, tubular flowers that are red when in bud and yellow at anthesis (Fig. 1). Individual plants produce 1 raceme with an average of 251.8 flowers (SD ± 92.0, N = 50). Kniphofia linearifolia occurs in populations restricted to wet grasslands and predominately flowers between February and April in KwaZulu–Natal. Each flower contains a dilute nectar reward (~12–15% sugar concentration) at the base of the tube (Brown et al. 2011). Flowers are protandrous, with each producing 6 anthers that are exposed for ~48 h after bud opening. After male phase, the stigma elongates and the flower enters female phase for up to 4–5 days. Kniphofia linearifolia is pollinated both by birds (sunbirds and weavers) and native bees (honeybees and solitary bees) (Brown et al. 2011). Previous work in two populations has shown that selective exclusion of birds can result in decreased fruit and seed set (Brown et al. 2011).
For this study, a total of 11 populations varying in size, density and isolation were chosen over approximately 40 km (Table 1; Fig. 2) near Pietermaritzburg, South Africa. These populations were selected to represent the natural range of population size, density and isolation that are characteristic of K. linearifolia (K.J. Duffy pers. obs.). Populations ranged in size from 37 to 1643 flowering individuals in 2010 and from 44 to 1361 flowering individuals in 2011 (Table 1).
Table 1. Characteristics of the 11 Kniphofia linearifolia populations sampled in this study in 2010, with 2011 values given in parenthesis
Total no. flowering plants
Density (plants m−2)
Karkloof Road B
Karkloof Road A
In 2010, 10 populations were each visited three times during the flowering period to quantify pollinator behaviour. On each visit, we made 15-min observations of bee foraging and 30-min observations of bird foraging on K. linearifolia. These time intervals were chosen as they represent a reasonable time period in which to observe at least one visit from each pollinator group. In 2010, a total of 22.5 h of pollinator observations were made among all populations, with a total of 7.5 h observations of bee visitation and a total of 15 h observations of bird visitation. For bee observations, in each period, we randomly selected one inflorescence in the population and counted the number of flowers visited by bees in 15 min. For bird observations, we randomly selected a patch of K. linearifolia plants in each population and stood at least 20-m distance from the patch so as not to disturb birds as they foraged. Patches observed for bird visitation varied in number of flowering individuals (range: 8–72 individuals) over the flowering season. We operationally defined population as a discrete group of flowering individuals that were separated by a minimum average of 1 km (calculated from three neighbouring populations) from other groups of flowering individuals. We defined patch as a group of flowering individuals within populations that were separated from other groups by at least 10 m in all populations, except in three small populations (Zeederberg Road, Murray Road, and Rotunda B), where patches were separated by at least 5 m. We recorded the number of visits made by observing head turns of birds between flowers using binoculars at 10 × magnification. We observed different plants in each observation period for both bee and bird observations. At the end of each observation period, we counted the number of birds foraging in the entire population as a measure of bird abundance for each population. We calculated the bee visitation rate per hour as the total number of bee visits × 4, as each plant was observed once for 15 min. We calculated the bird visitation rate per hour as the total number of bird visits × 2, as each patch was observed once for 30 min. All pollinator observations were made between 10:00 and 16:00.
Due to the non-independence of repeated pollinator observations within populations, we used generalised estimating equations (GEEs) (Liang & Zeger 1986) to examine the effects of population size, density and isolation on pollinator visitation to K. linearifolia. Population size was measured as the total number of flowering individuals in each population. Population density was calculated as the number of flowering plants divided by the area (m2) of the population. Area was calculated as the product of the maximum length and width of each population where K. linearifolia occurred using Global Positioning System (GPS) coordinates. Population isolation was measured by calculating the mean distance (km) of each population to each of the three nearest populations using GPS coordinates. These population parameters were all log-transformed prior to analyses to improve their fit. In addition, for the bird visitation rate analysis, we also included the number of plants observed as a measure of patch size. We tested whether these population parameters affected the abundance of birds, the hourly visitation rate by birds and the hourly visitation rate by bees separately using GEEs with a Poisson distribution incorporating a log link. In addition, we tested whether patch size varied according to population size, density or isolation using Pearson correlations. After log-transforming the variables, we found that patch size was positively related to population size (r =0.78; P < 0.001), but not population density (r =−0.32; P =0.08) or isolation (r =−0.04; P =0.833). We examined collinearity among fixed factors using variance inflation factor (VIF) values. For all fixed factors, VIF values were below 2.8, which is below the recommended upper threshold value of 3 (Zuur et al. 2009). For the GEE analyses, we used the geepack package for R (Halekoh, Højsgaard & Yan 2006). We included ‘population’ as a grouping factor in the models to account for the repeated sampling of populations and performed the analyses with an exchangeable correlation matrix and assessed significance using Wald statistics. We illustrate each of these independent effects graphically by isolating their effects on pollinator visitation from those of other influences included in the GEE output. To this end, we calculated log-transformed values for each independent effect and plotted these against the log-transformed response variable, with the appropriate curve generated from a generalised linear model between significant variables. All statistical analyses were performed with R (R Development Core Team 2012).
Fruit and Seed Set
At the end of the flowering period in 2010 and 2011, we estimated fruit set by counting both flowers that set fruit and flowers that did not set fruit on each plant for up to 50 plants per population. To estimate mean seed set per fruit, we sampled three fruits per plant, taken at different positions on the inflorescence. We also measured seed set per flower which was calculated as the mean number of seeds per fruit × proportion fruit set. Populations sampled in 2011 were the same as those sampled in 2010, apart from the inclusion of an additional small population in 2011 and the omission of three populations sampled in 2010 which could not be sampled in 2011 as they were either inaccessible or burned during the flowering period.
For both years' data, we tested whether population size, density and isolation affected fruit, seed set per fruit and seed set per flower. For this, we used population averages of fruit and seed set (Ågren 1996), because sampling was uneven among populations, and we did not want to bias the effect of population size for small populations. We used multiple regression for these analyses, with the predictor variables population size, density and isolation log-transformed prior to analysis. We tested for collinearity among predictors prior to analysis, using the methods described above, and found that this was within acceptable levels. Neither mean fruit nor seed set measures required transformation as they did not exhibit heteroscedasticity and residuals revealed that the models were a good fit to the data. In addition, we used regression to test for univariate relationships between bee and bird visitation rates as predictor variables and fruit set, mean seed number per fruit and mean seed number per flower as response variables. Both bee and bird visitation rates were log (x + 1)-transformed prior to analysis.
To quantify the contribution of bird pollination to fruit and seed set in each of 10 populations in 2010, we randomly selected five plants in bud (except for Rotunda B, where two plants were selected, due to this population flowering earlier than anticipated) and erected bird exclusion cages (20-mm mesh aperture diameter) over their inflorescences. To test whether these cages are effective in allowing insects to visit flowers freely while excluding birds from visiting flowers, we performed an experiment on the bee-pollinated plant Bulbine frutescens at the UKZN botanic garden. Native honeybees frequently visit B. frutescens. We erected the same exclusion cages used in the experiment on K. linearifolia on each of 10 patches, separated by at least 2 m, containing 4–5 flowering individuals of B. frutescens. We paired these caged individuals with uncaged plants that occurred adjacent to caged individuals. We quantified honeybee visitation for ten 15-min periods and found that caging has no effect on honeybee visitation rates (paired t test: t9 = 0.091; P =0.929), indicating that native honeybees can freely visit caged individuals. At the end of the flowering period, we counted fruit and seed set from a caged plant and an uncaged plant within 2–3 m that served as an unmanipulated control. We tested for the effect of caging treatment and population and their interaction using a two-factor anova on mean seed set per flower (mean proportion fruit set × mean seed set per fruit). We then tested whether population size, density and isolation affected seed set per flower in caged and uncaged plants by using separate ancova models with caging treatment as a fixed effect, and population size, density or isolation as a covariate. Due to earlier than expected fruit set, we could not estimate seed set from the Rotunda B and Cedara B populations in 2010 as the fruits had already dehisced.
To test whether fecundity in K. linearifolia was pollen-limited, in 2011, we added pollen from individuals from at least 5 m distance to each of 20 flowers on five separate plants in each of four populations (Bushwillow Park, Karkloof Bridge, Karkloof Road A, and Osborne Road). This was done by brushing pollen from a dehisced anther across the stigmas of recently opened flowers. At the end of flowering, we collected the fruits from supplemented plants and from open-pollinated plants within 2 m as controls. All pollen-supplemented flowers and their controls set fruit. We analysed the effect of supplementation on seed set per fruit using a two-factor anova with treatment and population and their interaction as fixed factors. Seed set data were log-transformed prior to analysis.
A total of 2132 bee visits per flower were observed over 7.5 h, and 2331 bird visits per flower were observed over 15 h, among all populations in 2010. Male and female Amethyst sunbirds (Chalcomitra amethystina) were the predominant bird species visiting K. linearifolia (86.4%), with occasional visits from Village weavers (Ploceus cucullatus) (12.1%) and White-bellied sunbirds (Cinnyris talatala) (1.5%). Native honeybees (Apis mellifera scutellata) were the predominant bee visitor (88.5%) with occasional visits by solitary bees, Hylaeus spp. (11.5%). On average, among all populations, the mean (± SE) rate of visitation by bees was 57.5 ± 13.9 visits per plant−1 h−1, while for birds it was 2.1 ± 0.5 visits plant−1 h−1. There was no effect of population size, density or isolation on bird visitation rate; however, bird visitation rate increased with increasing patch size (Table 2 and Fig. 3) indicating that birds are more active in larger populations. In addition, bird abundance increased with increasing population size (Fig. 4). There was no effect of population size and density on bee visitation rate to K. linearifolia individuals; however, bee visitation rate increased with increased population isolation (Table 2).
Table 2. Generalised estimating equation outputs for the effects of population parameters on the number of birds per population, bird visitation rate per hour and bee visitation rate per hour in populations of Kniphofia linearifolia in 2010
Bird visitation rate
Bee visitation rate
Patch size (log)
Population size (log)
Population density (log)
Population isolation (log)
Fruit and Seed Set
In 2010, the mean percentage of flowers that set fruit per plant ranged from 41.3 to 88.8% among populations and mean seed set per fruit per plant ranged from 0.67 to 29.0. Fruit set decreased with increasing population density and increased with increasing population isolation (Table 3; Fig. 5). Seed set per fruit increased with increasing population size, but not density or isolation (Table 3; Fig. 6). Univariate regressions revealed that fruit set (Fig. 7a) and mean seeds per flower (Fig. 7c) increased significantly with bird visitation rate. However, there was no significant relationship between bird visitation rate and mean seeds per fruit (Fig. 7b). There were no significant relationships between bee visitation rate and any of the measures of fecundity. In 2011, fruit set per plant ranged from 60.5 to 85.7% among populations and mean seed set per fruit per plant ranged from 0.00 to 32.33. There were no effects of population size, density or isolation on either fruit set or seed set in 2011 (Table 3).
Table 3. Multiple regression models for the effects of population parameters on fruit and seed production in Kniphofia linearifolia in 2010 and 2011. Analyses based on means of 10 populations in 2010 (3 and 6 d.f.) and eight populations in 2011 (3 and 4 d.f.)
Seed set per flower (F1,37 = 22.46, P <0.001; Fig. 8a) was reduced when birds were excluded from plants, indicating that birds are effective pollinators of K. linearifolia. This effect varied among populations (caging treatment × population interaction: F7,30 = 2.42, P =0.043). We found that there was a positive effect of population size on the difference in seed set per flower between caged and uncaged plants (caging treatment × population size interaction: F1,12=5.61, P =0.035; Fig. 8b). However, we did not find an effect of population density (F1,12=1.15; P =0.304) or population isolation (F1,12=0.07; P =0.797) on the difference in seed set per flower between caged and uncaged plants.
Seed set per fruit increased significantly when plants were hand-supplemented with cross pollen (back-transformed marginal means ± SE: 4.62; USE=0.52, LSE=0.46 for unmanipulated plants, vs. 7.24; USE=0.81, LSE=0.73 for pollen-supplemented plants; F1,30=8.66; P =0.006), indicating that fecundity in K. linearifolia populations may generally be pollen-limited. However, the effects of pollen supplementation did not vary significantly among the populations that received this treatment (treatment × population interaction: F3,30=0.857; P =0.474).
This study implicates birds as the causal agents of the positive relationship that we found between population size and fecundity in K. linearifolia (Fig. 6a). Birds, unlike bees, responded positively to two measures of aggregation, patch size and population size (Figs 3, 4 and Table 2), and they were also more effective than bees as pollinators (Fig. 8), despite being far less frequent as visitors. This finding contradicts recent claims that the abundance, rather than the per-visit effectiveness, of flower visitors best explains their contributions to fecundity (Vazquez, Morris & Jordano 2005). Nearly 90 per cent of observed bird visits were by Amethyst sunbirds; consequently, the likelihood of an Allee effect on fecundity in this species may depend on the availability of this particular bird species. Furthermore, we found significant positive relationships between bird visitation rate and both fruit set and seed set per flower (Fig. 7a,c). To our knowledge, this is the first study to attribute an Allee effect on fecundity of a plant to a particular minority subset of its flower visitors.
One reason why it has been difficult to attribute Allee effects to particular pollinators is because of temporal variability in pollinator visitation, due to factors such as weather conditions. This variation can often obscure underlying relationships between visitation rates and plant aggregation, particularly when observations must be made in different populations on different days and times. Nevertheless, we found that bird visitation rate was not affected by population size, but increased with increasing patch size (Table 2 and Fig. 3). The underlying reason for this positive association is probably that birds are resident for longer periods in larger populations. Indeed, we found that bird abundance increases with increasing population size. Sunbirds are itinerant and can travel longer distances between populations compared with bees; hence, they may also move to larger patches within populations or to larger populations to maximise their foraging efficiency. By contrast, there were no effects of population size or density on bee visitation rate; however, there was increased bee visitation with increased population isolation. This could reflect that bees are ubiquitous local residents and not limited by the floral resources in the K. linearifolia populations. However, they might forage more frequently in isolated populations to maximise their resource use, as there are longer distances to travel from these populations.
Although birds are more effective pollinators of K. linearifolia, bees also play a role in the reproduction of this species and do not necessarily always behave as ‘conditional parasites’ (sensu Thomson 2001) of this species. For instance, when we excluded birds from flowers we found that, on average, fruits were still set in 48.2% of flowers and these contained an average of 5.73 seeds. In related taxa (e.g. Aloe maculata), bees can be pollen thieves in some populations and primary pollinators in other populations, depending on the local availability of birds (Hargreaves, Harder & Johnson 2010; Duffy & Johnson 2011).
We found that fruit set was reduced in high-density populations, which may indicate either competition for pollinator attention or resources for fruit set with increasing density. However, fruit set may not be a reliable measure of fecundity in this species, as fruits can mature without containing fully developed seeds (K.J. Duffy, pers. obs.). As there is a putative late-acting self-incompatibility system in K. linearifolia (Duffy, Patrick & Johnson 2013), fruits can mature yet contain a full complement of aborted seeds (arising from pollinator-mediated self-pollination), or occasionally no seeds at all. Therefore, mean seed set per flower is the most reliable estimate of fecundity in K. linearifolia. Indeed, in 2010, we found that mean seed set per fruit per plant and mean seed set per flower increased with increasing population size, and not density or isolation, indicating the presence of an Allee effect that year. This is apparently due to the increased bird abundance in larger populations compared with smaller populations. We failed to detect a similar effect in 2011 probably because of differences in the populations examined and similarities in bird visitation rate between populations. This could be due to natural stochasticity in pollinator abundance and visitation rates between years among populations. Indeed, previous studies have demonstrated that component Allee effects can occur in 1 year and not in other years, which may be due to natural stochasticity in the relative abundance of plant populations and pollinators (Widen 1993; Ågren 1996; Alexandersson & Ågren 1996). It is yet to be quantified how long K. linearifolia individuals live. Observations on individuals from the UKZN botanic garden suggest that they can live for more than 25 years (A. Young, pers. comm.). Depending on resources and local environmental conditions, it is likely that individuals of K. linearifolia can potentially live many years in natural populations. Therefore, component Allee effects may occur in some years and not in others in this species, depending on the constitution of the local pollinator population and resources available between years. Although we demonstrate that K. linearifolia can suffer reduced fecundity in small populations, long-term analysis would be needed to quantify whether there are demographic Allee effects in K. linearifolia populations.
Rathcke (1983) proposed that fecundity would increase with increasing plant aggregation, of either conspecific or heterospecific individuals, to an asymptote and decline as pollinator availability becomes limiting in highly aggregated populations. This effect was not evident in our data, which showed that bird visitation accelerated with population size on the log scale, while seed set increased linearly. While increased aggregation has been shown to negatively affect per capita pollinator visitation in both plants with generalist pollinators (Duffy & Stout 2008) and specialist pollinators (Johnson, Hollens & Kuhlmann 2012), Rathcke's asymptote has seldom been demonstrated with field data (but see Brys, Jacquemyn & Hermy 2008). In generalist pollination systems, interspecific interactions among plants and flower visits may obscure the detection of Allee effects (Wagenius & Lyon 2010). In particular, fecundity of generalist plants can be enhanced by abundance of other species that attract pollinators (Johnson et al. 2003; Ghazoul 2006; but see Feldman 2008). Such increases in fecundity may be related to increased visitation rates and pollen transfer between conspecifics in the presence of increased heterospecific abundance (Duffy & Stout 2011). Although these ‘facilitation’ effects are not strictly Allee effects, which involve intraspecific as opposed to interspecific facilitation, they may shed light on whether the underlying basis for an Allee effect is mate or pollinator limitation. In our study, however, there were no other co-flowering bird-pollinated species that may have influenced bird foraging behaviour, and this may account for the positive relationships between K. linearifolia abundance and some measures of pollinator visitation and fecundity.
In conclusion, the type of pollinator may drive the mechanism underlying an Allee effect in a plant population. When populations fail to attract sufficient effective pollinators they can suffer from reduced per capita reproduction and become more reliant on other less effective pollinators. Such reliance on ineffective pollinators may leave such populations, regardless of size, prone to decreased fecundity and vulnerable to long-term decline.
We are grateful to Mark Brown for advice and help locating study populations and Alison Young for useful discussions. We thank two anonymous reviewers for valuable comments that greatly improved the quality of the paper. We thank Connal Eardley and Denis Brothers for help with solitary bee identification. This work was funded by a Postdoctoral Fellowship from the Claude Leon Foundation (K.J.D.) and a South African Research Chair (S.D.J.), which we gratefully acknowledge.