Author for correspondence: Maurice Paulissen Tel: +31 30 2536853 Fax: +31 30 2518366 Email: email@example.com
• In Dutch fens, subjected to high nitrogen (N) deposition, Scorpidium and other brown mosses have declined markedly. A concurrent strong increase of Sphagnum and Polytrichum has promoted acidification. We measured nitrate (NO3−) and ammonium (NH4+) availability in Dutch fens. We also tested preference for either N form of Scorpidium scorpioides, Sphagnum squarrosum and Polytrichum commune.
• Ion exchange membranes were installed in the field. In a hydroponic experiment, plants were grown on 100 µm N (reflecting concentrations in Dutch precipitation since 1980), provided as NO3−, NH4NO3, or NH4+.
• NH4+ availability in Sphagnum and Polytrichum stands and NH4+ : NO3− ratio in Sphagnum stands were higher than in brown moss stands. In the experiment, Scorpidium performed best on NO3−. NH4NO3 tended to decrease its growth, whereas NH4+ was very toxic. N treatment did not significantly affect growth of Sphagnum and Polytrichum. Tissue pH and nutrient concentrations confirmed the growth patterns and indicated that Scorpidium was most sensitive to NH4+ stress.
• We conclude that high NH4+ inputs pose a serious threat to the brown moss flora of rich fens.
Fens are minerotrophic peatlands, fed not only by precipitation (as in ombrotrophic peatlands), but also by groundwater or surface water. In contrast to poor fens, rich fen soils have a high base saturation (Sjörs, 1950), due to the influence of base-rich (but otherwise nutrient-poor) groundwater or surface water. The vegetation of pristine rich fens is very species-rich and contains many threatened species (Verhoeven & Bobbink, 2001). In terms of biomass and species diversity, bryophytes are very important in rich fens (Kooijman, 1992; Vitt, 2000). Brown mosses, including members of the genera Scorpidium, Calliergon and Campylium, dominate the bryoflora of rich fens (Gorham et al., 1987; Kooijman, 1992).
In the second half of the 20th century, many brown moss species have decreased markedly in Dutch rich fens, especially since 1980. Parallel to this decline, several Sphagnum and Polytrichum species have increased strongly (Kooijman, 1992; Paulissen, 2004). This has resulted in accelerated acidification of Dutch rich fens (turning them into poor fens) and in the development of floristically impoverished plant communities (Beltman et al., 1995; Schaminée et al., 1995).
Prolonged high N input into terrestrial ecosystems (including wetlands) can cause a shift in the dominant N species available to plant growth. This may change the species composition of plant communities (Bobbink et al., 1998). For example, prolonged loading of weakly buffered ecosystems with NH4+ initially stimulates nitrification, which generates protons. After the buffering capacity of the soil has been depleted, the soil pH starts to decrease and nitrification rate is strongly reduced. NH4+ then accumulates and the influx of NH4+ into plants increases (Bobbink et al., 1998). Although NH4+ uptake is more efficient than NO3− uptake in terms of energy cost, the presence of free NH4+ in plant cells is very toxic (Britto et al., 2001). Therefore, plants need to invest in a detoxification system in case the NH4+ uptake rate exceeds the rate at which it can be assimilated (Fangmeier et al., 1994; Krupa, 2003). Plant species show varying responses to increased external NH4+ concentration. Growth and survival of vascular plants from calcareous and weakly buffered habitats is generally reduced when the plants are transferred to an environment where NH4+ instead of NO3− is the dominant N form. By contrast, species from acid habitats generally perform better on soils where NH4+ is the sole N source or they are indifferent to N form (Gigon & Rorison, 1972; De Graaf et al., 1998; Lucassen et al., 2003; Dorland, 2004).
We assume that the typical brown moss flora of rich fens is optimally adapted to circumneutral habitats where part of the incoming NH4+ is converted to NO3−. We hypothesise that brown mosses prefer NO3− and that increased external NH4+ concentrations, as occurring in Dutch rich fens, are detrimental to these plants. By contrast, we hypothesise that NH4+ is the naturally dominant N species in poor fens (pH 4–5), where Sphagnum and Polytrichum dominate the bryophyte layer (Kooijman, 1993; Beltman et al., 1995). We also hypothesise that Sphagnum and Polytrichum, which replace brown mosses at a fast rate in Dutch fens, are less susceptible to increased external NH4+ concentrations. We assume that these species prefer NH4+ for growth.
In this paper, we present data on NO3− and NH4+ availability in Dutch fens as well as the results of a 3-month glasshouse trial. In this hydroponic experiment, we assessed the effects of different N forms (NO3−, NH4NO3 and NH4+, applied at an ecologically realistic concentration) on growth, tissue pH and nutrient status of three fen bryophyte species. The first, Scorpidium scorpioides (Hedw.) Limpr., is a brown moss species that has declined markedly in Dutch rich fens (Kooijman, 1992; Paulissen, 2004). The second species, Sphagnum squarrosum Crome, is a fast-growing species, which may directly replace S. scorpioides (Kooijman & Bakker, 1995) and has a strong acidifying capacity (Kooijman & Bakker, 1994). Finally, Polytrichum commune Hedw. has, since 1980, invaded many Dutch fens that are undergoing acidification (Paulissen, 2004). It establishes mainly on acid Sphagnum carpets, where it can rapidly become dominant.
Materials and Methods
Field measurements: inorganic N availability and pH
We measured NO3− and NH4+ availability and pH at 52 sampling points distributed over seven major Dutch fen reserves: Nieuwkoopse Plassen (52°9′-N, 4°49′-E), Het Hol (52°13′-N, 5°4′-E), Molenpolder (52°8′-N, 5°5′-E), Westbroekse Zodden (52°9′-N, 5°7′-E), Rottige Meenthe (52°49′-N, 5°54′-E), Weerribben (52°47′-N, 5°58′-E) and Wieden (52°41′-N, 6°6′-E). In each fen reserve, measurements were undertaken in three contrasting bryophyte vegetation types reflecting different succession stages: brown moss (including S. scorpioides), Sphagnum (including S. squarrosum) and Polytrichum (including P. commune).
At each sampling point, we installed two anion (AR204-SZRA) and two cation exchange membranes (CR67-HMR, Ionics Inc., Watertown, MA, USA). Optimal compatibility of the obtained anion and cation concentrations was assured by the following, strictly indiscriminate, protocol for the two membrane types. The individual membranes were cut to a size of 3 × 5 cm. Before installation, both membrane types were charged in a 1 m NaCl solution. The membranes were inserted in the bryophyte layer in such a way that the short side (3 cm) was perpendicular to the fen surface. The membranes were not visible after insertion. The top of the membranes was at a depth of about 2 cm (measured from the point where the bryophyte mat became compact). A first series of membranes was incubated for 2 months (June–August) in the summer of 2002. A second series was installed in August of the same year and removed 4 months later (autumn series). After removal from the bryophyte layer, the membranes were rinsed with demineralised water and carefully rubbed with latex gloves to remove any adhering plant and soil particles. The membranes were then briefly shaken to remove adhering drops of water, packed individually in small plastic bags and transported to the lab in a cool box. After storage at 4°C (maximum 1 wk), the ions bound to the membranes were extracted in 0.5 m NaCl for 4 h on a platform shaker. After storage at −20°C for 2 wk, NO3− and NH4+ concentrations in the extracts were measured on a Skalar SA-40 continuous-flow analyser (Skalar BV, Breda, the Netherlands). Summer and autumn data were averaged and the mean values were used for statistical analysis.
Water samples were taken from the top 10 cm of the bryophyte layer (measured from the point where the bryophyte mat became compact) in June, July, and August 2002, using Rhizon soil moisture samplers (Eijkelkamp BV, Giesbeek, the Netherlands). After storage at 4°C, pH of the samples was determined within 1 wk using a WTW 540 GLP pH meter with SenTix 41 electrode (Wissenschaftlich-Technische Werkstätten GmbH, Weilheim, Germany). For each sampling point, the values from the three consecutive months were averaged to obtain a mean pH value for statistical analysis.
Glasshouse experiment: design and measurements
Scorpidium scorpioides, Sphagnum squarrosum and Polytrichum commune were collected in De Stobbenribben (52°47′-N, 5°59′-E), a fen reserve in the Weerribben National Park. Model estimations of total N deposition in this part of the Netherlands amounted to 2.1–3.5 g N m−2 in 2001 (RIVM, 2003). The actual N deposition in De Stobbenribben is probably in the lower end of this range. On 20 February 2002, we collected S. scorpioides and S. squarrosum, while P. commune was sampled on 16 May 2002. The bryophytes were collected as turfs. In the glasshouse, the turfs were kept moist by applying tap water and a nutrient solution. The composition of this basic nutrient solution reflected average concentrations measured in Scorpidium, Sphagnum and Polytrichum stands in Dutch fens. It contained 200 µm HCO3−, 50 µm SO42–, 10 µm PO43–, 5 µm K+, 300 µm Ca2+, 50 µm Mg2+, 5 µm Fe3+ (plus 3.3 µm EDTA), 211.6 µm Na+, and 615 µm Cl−. HCl (1 m) was added to set pH at 5.5. This represents the average pH of the interstitial water measured at growth sites of the three species. The micronutrients Al, B, Co, Cu, Mn, Mo and Zn were added in very low concentrations (≤ 0.025 µm). Within 1 month, the top 5 cm of the bryophyte shoots were cut off and, after rinsing in tap water, distributed over perforated pots (∅ 5 cm × height 5 cm). Low nutrient concentrations and short exposure time ensured a minimal impact of the tap water on the bryophytes. Shoot densities in the perforated pots (1.8 cm−2 for S. scorpioides, 1.0 cm−2 for S. squarrosum, and 1.7 cm−2 for P. commune) were in range with those in the field. The pots were inserted into 0.5-l containers, which contained the basic nutrient solution. The containers were placed on a platform in a 4 × 1 × 0.5 m water bath, which was kept at 18°C to standardise growth conditions during the experiment.
After 2 wk of acclimatisation (during which P, K and Na supply was temporarily increased to 100, 50, and 256.6 µm, respectively, to minimise the risk of nutrient deficiency during the experiment), the basic nutrient solution was replaced by the different treatment solutions. These were prepared by adding one of three N forms to the basic nutrient solution: 100 µm NaNO3, 50 µm NH4NO3, 100 µm NH4Cl. This N concentration is in range with concentrations measured in Dutch wet deposition in the 1980s and 1990s (Paffen, 1990; Bobbink et al., 1992; Eerens et al., 2001; assuming average precipitation of 800 mm yr−1). During the experiment, nutrient solutions were replaced at least once, but mostly twice per week. The pH of the treatment solutions was set at 5.5 (i.e. the same value as in the basic nutrient solution). Analysis of the replaced growth medium showed that nitrification in the experimental containers was negligible (data not shown). The position of the experimental containers was randomised regularly. The fast-growing P. commune was harvested after 2 months, while S. scorpioides and S. squarrosum were harvested after 3 months.
At the end of the experiment, total shoot length (including side-branches) was measured. Next, the bryophyte shoots were rinsed with demineralised water to remove any attached drops of nutrient solution. The shoots were then carefully dried using paper tissues in a standardised way. Per replicate, total f. wt was determined 10 min after tissue application. The pH of the bryophyte tissue was determined by grinding one to four shoots per replicate (depending on shoot biomass) with liquid N2. The homogenised material was put on a Sentron 2001 pH electrode (Sentron BV, Roden, the Netherlands). Values were read as soon as the material had thawed (which was after about 10 s) and values had become stable. The plant material that had not been used for tissue pH determination was dried at 70°C for at least 72 h and total d. wt of the original replicate was calculated from the f. wt : d. wt ratio of this subsample. The dried material was ground using a ball mill and digested using a modified Kjeldahl method (Bremner & Mulvaney, 1982). The N, P and K concentrations of the diluted supernatant were determined colorimetrically (N and P) and flame-photometrically (K), using a Skalar SA-40 continuous-flow analyser. The Ca and Mg concentrations were measured on an IL type 357 atomic absorption spectrophotometer (Instrumentation Laboratory Inc., Wilmington, MA, USA).
Kruskal–Wallis tests followed by Dunn's posthoc analysis were carried out for each dependent variable and species, using succession stage (field data) and N treatment (experimental data) as a fixed factor, respectively. The Kruskal–Wallis tests were run in SPSS 8.0 (SPSS Inc., Chicago, IL, USA). Dunn's posthoc test is not included in SPSS 8.0 and was performed in MS Excel 2000 (calculation file copyright Edwin Martens, Centre for Biostatistics, Utrecht University). We chose these nonparametric tests because our data generally did not meet two important assumptions underlying parametric analysis of variance: homogeneity of variance and normality of error terms.
Field measurements in Dutch fens
NO3− availability was equal in all succession stages (Table 1). Compared to brown moss stands, NH4+ availability was more than 10 times higher in Sphagnum vegetation and about four times higher in Polytrichum stands. Total inorganic N availability did not differ significantly between the bryophyte vegetation types. The NH4+ : NO3− ratio was greater in Sphagnum stands than in brown moss vegetation. Polytrichum stands took in an intermediate position. The pH measured within the bryophyte stands decreased significantly as succession proceeded from brown moss via Sphagnum to Polytrichum (Table 1).
Table 1. Inorganic nitrogen (N) availability and pH in different succession stages in Dutch fens (subjected to high N deposition)
Available NO3− (µm in extract)
Available NH4+ (µm in extract)
TIN (µm in extract)
NH4+ : NO3− ratio
Total inorganic N (TIN) is the sum of nitrate (NO3−) and ammonium (NH4+). N availability was measured via ion exchange membranes. Rhizon soil moisture samplers were used to determine pH. Values are means ± 1 SE. Within each dependent variable, values sharing the same letter are not significantly different (P > 0.05).
Brown moss (n = 20)
5.4 ± 1.6a
5.0 ± 0.7a
10.3 ± 2.1a
3.6 ± 0.9a
6.3 ± 0.1a
Sphagnum (n = 20)
4.0 ± 1.5a
77.8 ± 29.7b
81.8 ± 30.7a
39.7 ± 20.1b
5.1 ± 0.1b
Polytrichum (n = 12)
7.1 ± 3.4a
19.3 ± 6.3b
26.3 ± 9.1a
5.4 ± 1.6ab
4.1 ± 0.1c
Length growth and d. wt at harvest Growth of S. scorpioides responded most markedly to variation in N form in the nutrient solution (Fig. 1a,b). The presence of NH4+, instead of only NO3−, led to a reduction in relative length increment of about three (NH4NO3 treatment) to four times (NH4+ treatment). However, only the difference between the NO3− and NH4+ treatments was significant. Compared to the NO3− treatment, S. scorpioides d. wt at harvest was significantly reduced in the NH4+ treatment. Cultivation on NH4NO3 resulted in intermediate mass growth. In contrast to S. scorpioides, N form did not significantly affect growth rate of S. squarrosum and P. commune (Fig. 1a,b).
Tissue pH In S. scorpioides, both the NH4NO3 and the NH4+ treatment led to a significant reduction in tissue pH compared to the NO3− treatment (Fig. 1c). In both treatments, the decline amounted to about 1 pH unit. By contrast, there was no effect of N form on tissue pH of S. squarrosum. The applied N form significantly affected tissue pH of P. commune. As in S. scorpioides, growth on solely NH4+ significantly reduced tissue pH (1 unit) compared to the NO3− treatment. However, unlike S. scorpioides, the negative effect of the NH4NO3 treatment on P. commune was much less pronounced (0.3 pH unit; Fig. 1c).
Tissue nutrient concentrations Tissue nutrient concentrations of the three species responded differently to variation in applied N form (Table 2). With the exception of Ca concentration, S. scorpioides consistently showed significant decreases in nutrient (including N) concentrations in the NH4NO3 and/or the NH4+ treatment, compared to the NO3− treatment. The decline was strongest in the case of P, K, and Mg, where the NO3− treatment resulted in tissue concentrations 2.5–3.7 times higher than the NH4+ treatment. N and Ca concentration showed less pronounced differences between treatments. In these cases, the NO3− treatment resulted in tissue concentrations of about 1.3 times higher than the lowest concentration (either the NH4NO3 or the NH4+ treatment). Differences between the NH4NO3 and the NH4+ treatments were consistently nonsignificant.
Table 2. Tissue nutrient concentrations (mg g−1 d. wt) as affected by nitrogen (N) treatment
Total applied N concentration was 100 µm in all treatments. Values are treatment means ± 1 SE; n = 5, except for P. commune, where n = 4 (NO3−), n = 2 (NH4NO3), and n = 3 (NH4+), respectively. Within each species and each dependent variable, values sharing the same letter are not significantly different (P > 0.05).
13.0 ± 0.5a
2.0 ± 0.1a
2.2 ± 0.3a
16.0 ± 2.4a
2.0 ± 0.1a
10.4 ± 0.2b
0.7 ± 0.0ab
0.7 ± 0.1b
16.0 ± 1.8a
1.2 ± 0.1ab
10.9 ± 0.4ab
0.6 ± 0.1b
0.6 ± 0.1b
13.2 ± 0.7a
0.8 ± 0.1b
7.4 ± 0.1m
0.6 ± 0.0m
1.6 ± 0.1m
10.0 ± 0.3m
0.9 ± 0.1m
8.2 ± 0.1mn
0.8 ± 0.0n
2.0 ± 0.1mn
6.9 ± 0.3mn
1.0 ± 0.1m
8.9 ± 0.0n
0.7 ± 0.0mn
2.0 ± 0.1n
4.6 ± 0.1n
0.6 ± 0.1m
10.2 ± 0.2x
0.7 ± 0.0x
2.9 ± 0.0x
6.7 ± 0.2x
0.9 ± 0.0x
10.4 ± 1.0x
0.7 ± 0.1x
3.0 ± 0.3x
6.7 ± 0.1x
1.0 ± 0.1x
11.1 ± 0.1x
0.7 ± 0.0x
3.0 ± 0.1x
3.7 ± 0.2x
0.6 ± 0.0x
N, P and K concentrations in S. squarrosum were about a factor of 1.3 lower in the NO3− treatment than in the NH4+-containing treatments. A different, and more pronounced, pattern was found for Ca. The concentration of this nutrient significantly decreased rather than increased with increasing dominance of NH4+ over NO3− in the nutrient solution. Tissue Mg concentration of S. squarrosum was not significantly affected by N treatment.
In P. commune, no significant effects of N form on nutrient concentration were found. However, Ca and Mg concentrations tended to be lower in the NH4+ treatment than in the NO3− and NH4NO3 treatments (Table 2).
This study presents data on NO3− and NH4+ availability as measured with ion exchange membranes in Dutch fens under high N deposition. It also provides insight into the effects of different forms of inorganic N on growth and physiological status of typical bryophytes from different succession stages in fens.
Our data on inorganic N availability in Dutch fens indicate that much of the incoming NH4+ in brown moss stands is nitrified, due to the circumneutral pH. Consequently, part of the NO3− that is formed may be lost from the ecosystem by denitrification, as suggested by equal NO3− availability in the three succession stages. Indeed, the main effect of ‘succession stage’ on total inorganic N availability (P = 0.08; Kruskal–Wallis test) indicated that this variable tended to be lowest in brown moss vegetation. Even if nitrification is high in brown moss stands, increased atmospheric NH4+ input is likely to be problematic for species such as S. scorpioides, as the emergent shoot tips of brown mosses are directly exposed to high NH4+ concentrations in precipitation. Moreover, despite the occurrence of nitrification, NH4+ availability in Dutch brown moss stands was significantly higher than in brown moss stands in a low deposition site in central Ireland (data not shown). NH4+ : NO3− ratio suggested that nitrification also occurred in Polytrichum stands. This was not expected, but De Boer & Kowalchuk (2001) have shown that nitrification can occur even at low pH. Nitrification was probably promoted by higher O2 availability in the Polytrichum mats, which were mostly somewhat drier than both brown moss and Sphagnum stands.
None of the fen bryophyte species that we tested performed better in the NH4+ treatment than in the NO3− or the mixed treatments. However, there were marked differences in sensitivity to NH4+. Our data show that, when exposed to N concentrations realistic for wet deposition in the Netherlands (cf. Paffen, 1990; Bobbink et al., 1992; Eerens et al., 2001), S. scorpioides is much more sensitive to increased external NH4+ concentrations than both S. squarrosum and P. commune.
In general, our data correspond well with the typical effects of NH4+ toxicity as found in the above-mentioned studies (Gigon & Rorison, 1972; Soares & Pearson, 1997; De Graaf et al., 1998; Krupa, 2003; Lucassen et al., 2003; Pearce et al., 2003). However, with increasing external NH4+ concentration, tissue N and P concentrations of S. scorpioides showed a significant decrease rather than an increase. In fact, all nutrients that we measured (except Ca) showed a strong decline in this species. This underlines the strong toxic effect of NH4+ on S. scorpioides. Taking into account the sharp growth reduction, resulting from NH4+ stress, we assume that S. scorpioides suffers from severe membrane dysfunction (cf. Fangmeier et al., 1994; Krupa, 2003) and cell solute leakage, as previously described for the N-sensitive bryophyte Racomitrium lanuginosum (Pearce et al., 2003). In S. squarrosum, there was a tendency towards lower mass growth in the NH4+ treatment. This probably contributed to the unexpected increase in tissue K concentration in this species under elevated external NH4+ concentrations (K content (mg) suggested that K uptake was not increased in the NH4+ treatment; data not shown). In S. squarrosum, NH4+ stress was obviously relatively low since tissue K and Mg concentrations did not indicate significant leaching of these elements.
In addition to the physiological responses described above, it has been shown in vascular plants that growth on NH4+ instead of NO3− mostly leads to a decrease in tissue pH (Gerendás et al., 1990; De Graaf, 2000; Lucassen et al., 2003). This phenomenon is linked to the assimilation of NH4+ into amino groups, which produces H+ (Raven & Smith, 1976). The plant has to excrete or neutralise this H+ in order to maintain cell pH homeostasis. Our data show that the NH4+-insensitive S. squarrosum is better capable of maintaining constant tissue pH than S. scorpioides and, to a lesser extent, P. commune. However, growth and tissue nutrient response of P. commune showed that a decrease in tissue pH did not cause serious physiological problems in this species. A negative correlation between external NH4+ concentration and tissue pH has also been found in both NH4+-insensitive and sensitive vascular plant species (Dorland, 2004). This indicates the limited value of plant tissue pH as an indicator of NH4+ stress.
How can the results of our short-term glasshouse trial be translated to the field situation in Dutch rich fens? We have shown that, as Sphagnum and Polytrichum invade, rich fens experience a gradual pH decline resulting in values as low as 4 (cf. Kooijman, 1993; Beltman et al., 1995). Under such acid conditions, NH4+ is more toxic to plants than at higher pH (> 5), as has been shown for vascular plants from calcareous and weakly buffered habitats (Gigon & Rorison, 1972; Lucassen et al., 2003; Dorland, 2004). Moreover, our data indicate that at lower pH nitrification is inhibited in the wet Sphagnum stands, which causes NH4+ accumulation. This is in agreement with a previous study (Kooijman, 1993). The combination of lowered pH and increased external NH4+ concentrations, as experienced by brown mosses neighbouring fast-growing Sphagna, will considerably reduce brown moss growth rate and competitive strength. By contrast, in our experiments growth of S. squarrosum and P. commune was not or only slightly affected by high external NH4+ concentrations. Considering the vigorous growth of invasive Sphagnum and Polytrichum species in Dutch fens, it is likely that they are similarly insensitive to NH4+ under more acid conditions in the field. It may be that these acidophytic bryophytes possess a well-developed NH4+ detoxification system (e.g. incorporation in N-rich amino acids: cf. Näsholm et al., 1994; Limpens & Berendse, 2003). Such a detoxification mechanism may be much less developed in brown mosses, which are, in undisturbed habitats, subjected to low external NH4+ concentrations and comparatively high concentrations of nontoxic NO3−.
We conclude that an increase of external NH4+ concentrations due to high atmospheric deposition is especially detrimental to the rich fen bryophyte S. scorpioides. However, it does not significantly affect the successor species S. squarrosum and P. commune. This clearly indicates that NH4+ toxicity may play an important role in the decline of typical rich fen brown mosses in the Netherlands. Additional field experiments should validate these results for the longer term.
We thank Natuurmonumenten and Staatsbosbeheer for fieldwork allowances. In particular, we are grateful to Mr J. Bredenbeek (Staatsbosbeheer Flevoland-Overijssel) for permission to collect bryophytes in the Stobbenribben nature reserve. We also thank Laura Espasa Besalú, Sandra Robat, Gerrit Rouwenhorst and Deirdre Scholts for their help during the experiment and sample processing. Dr Sylvia Toet provided us with the ion exchange membrane protocol and Edwin Martens (Centre for Biostatistics, Utrecht University) gave statistical advice. Finally, we thank Prof. Jos Verhoeven, Prof. Ian Woodward, and three anonymous referees for their valuable comments on earlier drafts of the manuscript.