Behaviour of chemical elements during weathering of pyroclastic rocks, Hong Kong

Environment International 26 (2001) 359 ± 368

www.elsevier.com/locate/envint

Behaviour of chemical elements during weathering of pyroclastic rocks, Hong Kong J. Malpas, N.S. Duzgoren-Aydin*, A. Aydin Department of Earth Sciences, University of Hong Kong, Pokfulam Road, Hong Kong SAR, People's Republic of China

Abstract The behaviour of whole-rock major, trace and rare earth elements (REE) during weathering under subtropical conditions is examined along a profile developed over crystal ± vitric tuffs with eutaxitic texture. The intensity of weathering within the profile varies erratically, indicating weathering processes operate over different scales. Quartz, K-feldspar, plagioclase and biotite are the main primary minerals, whereas clays, sesquioxides, sericite and chlorite are the alteration products. Kaolinite, halloysite and illite-mica are the dominant clay minerals present in significantly varying proportions. Two competing processes, namely leaching and fixation, are the main regulators of variations in mostly major and some trace element concentrations along the profile. In general, as the intensity of weathering increases, Ca, Na, K, Sr ‹ Si decrease, while Fe, Ti, Al and loss of ignition (LOI) increase. Likewise, the intensity of negative Eu-anomaly decreases while the intensity of negative Ce-anomaly and the La/Lu and Sm/Nd ratios increases. In detail, however, the behaviour of chemical elements cannot be solely explained in terms of the degree of weathering. This study makes it clearly evident that the type and abundance of sesquioxides and clay minerals can significantly modify the geochemical signatures of weathering processes. D 2001 Elsevier Science Ltd. All rights reserved. Keywords: Clay minerals; Geochemistry; Hong Kong; Pyroclastic rocks; Weathering

1. Introduction The behaviour of chemical elements during weathering and the nature of fluid/rock interactions were the focus of many recent studies (Nesbitt, 1979; Gouveia et al., 1993; Ohlender et al., 1996; Minarik et al., 1998; Tapia et al., 1998). The behaviour of elements during alteration mainly depends on the relative stability of the parent mineralogy, pressure, temperature, redox potential and the type of leaching agent. Mobility of chemical elements is conventionally presented by percentage changes in ratios of elements from weathered samples, relative to those from fresh (parent) rock. These ratios are calculated by dividing each element by one of the so-called ``immobile'' elements. However, there are growing number of studies suggesting that these ``immobile'' elements which include rare earth elements (REE), Ti and Al, may be mobile during weathering (Gardner, 1980; Gardner et al., 1978; Van Der Weijden and Van Der Weijden, 1995; Cornu et al., 1999).

* Corresponding author. Tel.: +852-2964-5799; fax: +852-2517-6912. E-mail address: [email protected] (N.S. Duzgoren-Aydin).

In engineering geological studies, variations in chemical elements along weathered profiles were utilised to predict changes in the degree of weathering (Ruxton, 1968; CostaFilho et al., 1989). Many types of chemical indices based on the percentages of various combinations of major oxides were formulated to capture these changes (Reiche, 1943; Rocha-Filho et al., 1985; Sueoka, 1988). However, none of these studies considered the potential effects of sesquioxides and clay minerals on the behaviour of chemical elements. This paper examines the distribution and mobility of whole rock chemical elements during the weathering of pyroclastic rocks under subtropical conditions and gives special emphasis to the type and the abundance of sesquioxides and clay minerals. This study forms part of a wider investigation aimed at reassessing the chemical weathering indices for the characterisation of profile heterogeneities in engineering geological applications. 2. Materials and methods The nine samples collected represent a profile derived from pyroclastic rocks weathered under subtropical condi-

0160-4120/01/$ ± see front matter D 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 1 6 0 - 4 1 2 0 ( 0 1 ) 0 0 0 1 3 - 7

360

J. Malpas et al. / Environment International 26 (2001) 359±368

tions. The profile forms a section of a well-drained mountainous terrain typical of Hong Kong's landscape. A six-fold weathering grade scheme (BSI, 1990) was adopted to classify the degree of weathering of the samples. Samples from sesquioxide-rich zones were classified as completely weathered (i.e., Grade V) and labelled as VsqL, VsqM, VsqD, on the basis of visual assessment of their sesquioxide content. Sample VsqL is almost free of sesquioxides and light in colour; VsqM contains intermediate amounts of sesquioxides and has a darker colour than VsqL; and VsqD has a significant amount of sesquioxides and is the darkest in colour. In this study, atomic absorption spectrometry (AAS), Xray fluorescence spectrometry (XRF) and inductively coupled plasma-mass spectrometry (ICP-MS) were utilised for whole rock major, trace and REE analyses, respectively. Petrographic examinations were carried out using resin-impregnated polished thin sections under a polarised light optical microscope. Scanning electron microscopy (SEM) combined with energy dispersive X-ray analyser (EDAX) was employed for semiquantitative chemical analyses of sesquioxides. The thin sections used for optical examination were also employed for EDAX analyses after carbon coating. X-ray diffraction (XRD) and optical microscopy were employed for mineral identification. XRD analysis was carried out using Philips PW3040 series diffractometer equipped with a copper-target tube, operating at 50 kV and 40 mA. Samples were air-dried, the < 2 mm fractions were dispersed in 0.01 M hexametaphosphate, and subsequently the < 2 mm fraction were collected for clay mineral analysis. XRD patterns of clay fractions coated onto ceramic tiles were gathered after various pretreatments to enable the identification of clay types (Brown and Brindley, 1980). For each sample, three tiles were prepared: one was used to differentiate 1:1 type clay minerals, particularly halloysite from kaolinite by the formamide intercalation method (Churchman et al., 1984); the other

two, K- and Mg-salt saturated tiles, which had been subjected to subsequent heat and glycerol treatments, respectively, were used to differentiate 2:1 clay types (Brown and Brindley, 1980). 3. Results and discussion 3.1. Petrography and mineralogy Fresh (Grade I) samples, crystal ± vitric tuffs with eutaxitic texture, contain subhedral to euhedral quartz, K-feldspar, plagioclase and biotite grains, few lithic fragments and a crystal ± vitric groundmass (Fig. 1). Quartz and Kfeldspar show no visible evidence of alteration, whereas plagioclases exhibit slight sericitisation. Biotite grains, on the other hand, are moderately to strongly altered to chlorite. Chlorite is restricted to fresh samples, whereas sericite is found throughout the weathered profile due to alteration of plagioclase as well as K-feldspar. During petrographic examinations, the term ``sericite'' is adopted to describe micaceous alteration products as defined by Millot (1970) and corresponds to illite-mica as identified by XRD. In Grade III samples, chlorite grains are completely replaced by clay and sesquioxides. A mixture of sericite ‹ clay ‹ sesquioxides replaces groundmass as well as plagioclase grains. The replacement is generally pseudomorphic implying that initial shape, bulk volume and/or internal texture of the replaced mineral are preserved (Merino et al., 1993). Outlines of the primary grains are traced by Fe-staining. K-feldspar grains display slight to moderately developed sericitisation. Relatively continuous inter- and intragrain fractures are common. Quartz, Kfeldspar and plagioclase grains within the same thin section are significantly different in their degree of weathering, in mineral-scale. Such observations are well documented and clearly reflect that different minerals have different degree

Fig. 1. Optical photomicrography of SW-3 representing parent rock (Grade I). Field of view is 1.5 mm. Plain-polarised light.

J. Malpas et al. / Environment International 26 (2001) 359±368

361

Fig. 2. Optical photomicrography of relict primary grain (plagioclase ?) replacing pseudomorphically by a mixture of sericite, clay and sesquioxides. Field of view is 1.5 mm. Crossed-polarised light.

of resistance to weathering (e.g., Goldich, 1939; Laughman, 1969). It is also not uncommon to observe that the degree of weathering of the same mineral varies from slight to moderate, even in the same thin section, suggesting chemical disequilibrium. At more advanced stages (Grade IV ± V) almost all primary minerals, except quartz, are replaced by a

mixture of secondary minerals, mostly composed of sesquioxides and clay minerals with a minor amount of sericite. Quartz, as the most resistant mineral, begins to reveal some evidence of dissolution, particularly along the grain boundaries. Most of the grains are strongly fractured. The amount of sesquioxides and clay minerals increases significantly with the degree of weathering

Fig. 3. Oriented clay XRD patterns of SW-2 (Grade III) with the formamide and heat treatments. Order of abundance of clay minerals is illitemica > halloysite>kaolinite. I-M: illite-mica; H.H: hydrated halloysite; K: kaolinite; F-110 C: heat treatment.

362

J. Malpas et al. / Environment International 26 (2001) 359±368

(Fig. 2), and in places, clay minerals are themselves replaced by sesquioxides. Sesquioxides are ubiquitous throughout the profile and have close textural and mineralogical associations with clay minerals. Samples from the sesquioxide-rich zone (VsqL, VsqM and VsqD) display a network of sesquioxide veins as a characteristic microfabric feature. These veins are typically rich in Mn and contain Fe, Pb, Ba and Ce in variable proportions. Similar to other Grade V samples, samples from the sesquioxide-rich zone are dominated by quartz and secondary minerals. In addition to their high content of sesquioxides, the abundance of sericite is also notably high. Kaolinite, halloysite, illite-mica and chlorite are identified by XRD. Chlorite is identified only in the fresh sample. Although illite-mica is common along the weathered profile, it is the dominant clay type only in Grade III and better samples (Fig. 3). As the degree of weathering increases, dominant clay mineralogy changes from illitemica to kaolin-group minerals, including kaolinite and halloysite (Fig. 4). The relative abundance ratio of kaolinite to halloysite varies noticeably along the weathered profile. Halloysite is the dominant kaolin mineral in

samples Grade III and Grade III ±IV, whereas kaolinite dominates the clay mineralogy in Grade IV and worse samples. Samples from the sesquioxide-rich zone have higher kaolinite and illite-mica contents compared with other samples with the same degree of weathering but lower sesquioxide content. 3.2. Geochemistry In this study, no element has been assumed to be immobile or used as the denominator. Instead, all elements are normalised using their parent rock concentrations (Xsample/Xparent). If the normalised value is less than 1, it implies that element X of the weathered sample has been depleted (mobilised) with reference to its parent rock. Conversely, if the value is higher than 1, it suggests that element X has been enriched (fixated). Note that parent-normalised diagrams are logarithmic in scale (Figs. 5 and 6). Bulk rock major, trace and REE data are tabulated in Tables 1 and 2. The effects of weathering on the major and minor element concentrations and loss of ignition (LOI) are illustrated in Figs. 5 and 6. In general, as the intensity of

Fig. 4. Oriented clay XRD patterns of SW-1.4 (Grade V) with the formamide and heat treatments. Order of abundance of clay minerals is kaolinite>halloysite>illite-mica. I-M: illite-mica; H.H: hydrated halloysite; K: kaolinite; F-110 C: heat treatment.

J. Malpas et al. / Environment International 26 (2001) 359±368

363

Fig. 5. Parent-normalised major and minor oxides versus weathering grade.

weathering increases, Ca, Na, K, Sr ‹ Si decrease, whereas LOI, Fe, Ti and Al increase. Depletions of Ca, Na, Sr and K are closely controlled by alteration of feldspar. This is consistent with the petrographic observations suggesting that feldspar is more prone to weathering than quartz. Relative enrichment of LOI, Fe, Ti and Al can be attributed to the formation of secondary phases, particularly sesquioxides and clay minerals. These observations suggest that two competing processes (leaching and fixation) control mostly major and some trace element concentrations during weathering. In detail, however, the behaviour of chemical elements cannot be solely explained in terms of weathering grade. It is believed that type and abundance of clay minerals might significantly modify the behaviour of chemical elements, particularly K and LOI. For example: 

The moderately decomposed sample (SW-2; Grade III) has moderately to strongly altered feldspar



grains, but it has a K2O content very similar to the parent rock (SW-3; Grade I). This can be explained in terms of significant abundance of illite-mica (Kbearing clay). The moderately to strongly altered sample (SW-1.3; Grade III ±IV) has the highest LOI and halloysite content. This sample also has the highest Cs. This can be attributed to the effects of organic compounds (e.g., Staunton and Levacic, 1999) and/or the clay type on Cs absorption.

As illustrated in Figs. 5 and 6, samples VsqM and VsqD display significantly greater enrichment in Mn, Fe, Mg, Pb, Ba, Cu, Cr, Sc and Sr compared to samples (SW-1 ±4 and/or VsqL) displaying the same degree of weathering but containing lower sesquioxide content. Additionally, the sesquioxide-rich samples (VsqM and VsqD) have slightly higher K, Ca and Na contents compared to the sesquioxide-poor samples (SW-1± 4 and

364

J. Malpas et al. / Environment International 26 (2001) 359±368

Fig. 6. Parent-normalised trace elements versus weathering grade.

VsqL). It is thus evident that the abundance of sesquioxides significantly modifies the distribution of mostly trace and some major elements. The effects of weathering on REE concentrations are illustrated in Fig. 7 where REE data are normalised to chondrite and parent rock abundances, respectively. Excluding the sample VsqM (representing samples from the sesquioxide-rich zone), all samples display similar REE patterns. Chondrite-normalised parent rock (Grade I) exhibits light REE (LREE) enrichment relative to heavy REE (HREE) with a significant negative Eu-anomaly that can be attributed to plagioclase fractionation. As the intensity of weathering increases, total REE abundances and the intensity of the negative Ce-anomaly increase, while the intensity of the negative Eu-anomaly decreases (Figs. 7 and 8). Note that the parent rock displays neither

a positive nor a negative Ce-anomaly. Similarly, as the weathering advances, the La/Lu and Sm/Nd ratios increase (Fig. 8). On the other hand, the sample from the sesquioxide-rich zone (VsqM) displays a zigzag pattern and is characterised by middle REE (MREE) enrichment relative to HREE with negative Ce-anomaly and positive Eu-anomaly. It is also apparent that sesquioxides modifies the distribution of REE. 4. Conclusions The distribution and mobility of chemical elements during weathering of crystal ±vitric tuff under subtropical conditions were examined. Two competing processes, namely leaching and fixation, regulate variations in chem-

J. Malpas et al. / Environment International 26 (2001) 359±368

365

Table 1 Whole rock major and trace element data Weathering grade

I

III

III ± IV

IV ± V

IV ± V

V

VsqL

VsqM

VsqD

Sample number

SW-3

SW-2

SW-1.3

SW-1.1

SW-1.2

SW-1.4

SW-1.5L

SW-1.5M

SW-1.5D

75.27 0.06 11.78 1.80 0.00 0.01 0.14 0.12 2.06 0.01 8.32 99.57

75.01 0.12 15.78 1.10 0.04 0.04 0.21 0.15 2.02 0.02 5.96 100.45

72.24 0.08 17.12 1.04 0.03 0.04 0.10 0.11 2.12 0.01 7.06 99.95

72.41 0.04 17.64 1.19 0.06 0.05 0.10 0.11 1.57 0.01 6.40 99.58

75.15 0.05 15.21 1.49 0.10 0.06 0.10 0.09 1.31 0.02 5.96 99.54

71.85 0.04 11.32 4.68 3.87 0.08 0.15 0.11 1.65 0.01 5.89 99.65

70.68 0.04 9.46 7.35 4.73 0.10 0.17 0.11 1.99 0.02 5.64 100.29

Whole rock (wt.%) SiO2 TiO2 Al2O3 Fe2O3 MnO MgO CaO Na2O K2O P2O5 LOI Total Trace element (ppm) Sc V Cr Ni Cu Zn Rb Sr Y Zr Nb Ba Ce Pb Th U Cs

77.70 0.02 12.17 1.04 0.05 0.03 0.36 3.16 4.86 0.00 0.53 99.92 2 29 0 2 ld 306 269 25 65 182 23 86 119 42 41 21 3

74.19 0.07 14.83 1.04 0.05 0.11 0.14 0.17 4.35 0.01 4.62 99.58 ld 30 ld ld ld 55 325 5 52 244 31 84 44 27 41 2 6

ld 30 ld ld 5 80 244 0 58 259 35 45 47 60 57 0 22

ld 29 ld ld 1 34 143 0 108 246 28 57 84 71 44 3 4

ld 33 ld ld 2 41 147 1 124 232 30 52 122 55 50 2 3

ld 31 13 ld 2 48 114 ld 175 186 19 43 154 192 48 ld 6

ld 26 ld ld 4 52 120 ld 164 179 19 108 213 152 28 ld na

13 ld 356 10 45 110 131 4 203 187 5 2228 6142 3571 108 21 na

18 ld 53 12 63 152 154 11 211 198 1 2947 6502 5547 135 28 na

Explanations: na, not applicable; ld, less than detection limit.

ical element concentrations along the profile. In general, during weathering, depletion of certain elements such as Ca, Na, K and Sr can be explained in terms of the

breakdown of primary minerals, particularly plagioclase and K-feldspar, while the enrichment of LOI, Fe, Ti, and Al can be attributed to the formation of clay minerals and

Table 2 Whole rock REE data (ppm) Weathering grade

I

III

III ± V

IV ± V

IV ± V

V

Vsq

Sample number

SW-3

SW-2

SW-1.3

SW-1.1

SW-1.2

SW-1.4

SW-1.5M

Chondrite

La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu

47.43 97.39 11.80 45.70 10.02 0.18 8.83 1.44 9.39 1.88 5.67 0.85 6.34 0.89

44.50 48.88 10.77 42.00 8.51 0.18 7.47 1.17 7.96 1.71 5.48 0.82 6.37 0.87

73.03 55.01 18.10 70.35 14.51 0.40 10.97 1.66 10.15 1.88 5.59 0.83 6.32 0.84

164.97 59.01 35.80 128.27 24.63 0.72 16.53 2.40 15.57 3.12 9.09 1.35 9.57 1.26

217.96 96.22 45.45 216.72 31.72 0.89 20.14 2.86 18.01 3.48 10.12 1.43 10.08 1.34

305.86 193.78 80.10 310.08 50.98 1.57 32.87 4.67 27.13 5.45 15.42 2.22 14.63 2.03

277.13 93.85 54.92 ÿ 66.78 31.16 1.54 3.67 2.15 32.31 8.53 30.61 2.25 16.28 2.32

0.33 0.87 0.13 0.63 0.20 0.08 0.28 0.05 0.34 0.08 0.23 0.04 0.22 0.03

Chondrite REE values are from Sun (1982)

366

J. Malpas et al. / Environment International 26 (2001) 359±368

Fig. 7. Chondrite- and parent-normalised REE diagrams.

sesquioxides. However, this study demonstrates that not only the degree of weathering, but also the abundance of clay minerals and sesquioxides control the behaviour and mobility of chemical elements. More specifically, abundance of halloysite can significantly modify LOI and Cs contents, while illite-mica can affect the K and/or Sr contents of the samples. Mn-rich sesquioxides closely controls the enrichment of several elements including Fe, Mg, K, Na, Ca, Ba, Pb Cu, Ni and Sr. Such enrichment can be explained in terms of high sorption affinity of Mn-oxides, particularly for metal cations. Additionally, significant enrichment of Ba and Sr together with low valance cations including Mg, Cu and Ni may suggest the presence of todorokite (a-MnO2; Kampf et al., 2000). Similarly, the presence of hollandite (Ba-rich aMnO2) and/or coronadite (Pb-rich a-MnO2) can also

explain the high content of trace elements of the samples from the zone. However, except for vernadite (d-MnO2), the presence of other Mn-minerals has not been confirmed by XRD. This can be attributed to the fact that except for vernadite, other Mn-minerals are not sufficient enough to be identified by XRD, or alternatively, sesquioxides mostly occur as amorphous coating material. The distribution of REE patterns along the profile indicates that REE have been fractionated during weathering. Fractionation of REE can be explained in terms of alteration of primary minerals and progressively advancing oxidising conditions (cf. Nesbitt, 1979). However, it is the abundance of sesquioxides that mainly modifies the distribution of REE and that suggests the presence of different chemical environments, such as pH and leaching conditions, and the mineralogical control.

J. Malpas et al. / Environment International 26 (2001) 359±368

367

Fig. 8. Chondrite-normalised REE ratios versus weathering grade.

There are three main factors that control the behaviour of chemical elements during weathering: the degree of weathering, the type and the abundance of sesquioxides and clay minerals. Elemental geochemistry of weathered samples can provide significant information about the nature of fluid/ rock interaction during weathering process. More importantly, this study clearly emphasises the necessity to reassess the usage of chemical weathering indices in the characterisation of weathered profiles for engineering purposes. Acknowledgments This research has been founded by CRCG, University of Hong Kong. References British Standards Institution. BS 1377, British standards methods of test for soil for civil engineering purposes, 1990. Brown G, Brindley GW. X-ray diffraction procedures for clay mineral identification. In: Brindley GW, Brown G, editors. Crystal structures of clay minerals and their X-ray identification. London: Mineralogical Society, 1980. pp. 305 ± 59. Churchman CJ, Whitton JS, Claridge GGC, Theng BKG. Intercalation method using formamide for differentiating halloysite from kaolinite. Clays Clay Miner 1984;32:241 ± 8. Cornu S, Luca Y, Lebon E, Ambrosi JP, Luizao F, Rouiller J, Bonnay M, Neal C. Evidence of titanium mobility in soil profiles, Manaus, central Amazonia. Geoderma 1999;91:281 ± 95. Costa-Filho LM, Dobereiner L, de Campos TMP, Vargas E. Fabric and engineering properties of saprolites and laterites. Twelfth Int Conf SMFE, Rio de Janeiro 1989;4:2476 ± 763.

Gardner LR. Mobilization of Al and Ti during weathering Ð isovolumetric geochemical evidence. Chem Geol 1980;30:151 ± 65. Gardner LR, Kheoruenromne I, Chen HS. Isovolumetric geochemical investigation of a buried granite saprolite near Columbia, SC, USA. Geochim Cosmochim Acta 1978;42:417 ± 24. Goldich SS. A study in rock weathering. J Geol 1939;46:17 ± 58. Gouveia MA, Prudencio MI, Figueiredo MO, Pereria LCJ, Waerenborgh I, Mprgado I, Pena T, Lopes A. Behaviour of REE and other trace and major elements during weathering of granitic rocks, Evora, Portugal. Chem Geol 1993;107:293 ± 6. Kampf N, Scheinost AC, Schulze DC. Oxide minerals. In: Summer ME, editor. Handbook of soil science. CRC Press, 2000 (F-125, F-168). Laughman FC. Chemical weathering of silicate minerals Elsevier: New York, 1969. Merino E, Nahon D, Wang Y. Kinetics and mass transfer of pseudomorphic replacement of parent minerals and kaolinite by Al, Fe, and Mn oxides during weathering. Am J Sci 1993;293:135 ± 55. Millot G. Geology of clays New York, NY: Springer-Verlag, 1970. Minarik L, Zigova A, Skrian P, Stastny M. The behaviour of rare-earth elements and Y during the rock weathering and soil formation in the Ricany granite massif, Central Bohemia. Sci Total Environ 1998;215: 101 ± 11. Nesbitt HW. Mobility and fractionation of rare earth elements during weathering of granodiorite. Nature 1979;279:206 ± 10. Ohlender B, Land M, Ingri I, Widerlund A. Mobility of rare earth elements during weathering of till in northern Sweden. Appl Geochem 1996; 11:93 ± 9. Reiche P. Graphic representation of chemical weathering. J Sediment Petrol 1943;13(2):58 ± 68. Rocha-Filho P, Antuenes FS, Falcao MFG. Quantitative influence of the weathering upon the mechanical properties of a young gneiss residual soil. First Int Conf on Geomechanics in Tropical Lateritic and Saprolitic Soils, Brasilia 1985;1:281 ± 94. Ruxton BP. Measures of the degree of chemical weathering of rocks. J Geol 1968;76:518 ± 27. Staunton S, Levacic P. Cs adsorption on the clay-sized fraction of various soils: effect of organic matter destruction and charge compensating cation. J Environ Radioact 1999;45:161 ± 72.

368

J. Malpas et al. / Environment International 26 (2001) 359±368

Sueoka T. Identification and classification of granitic residual soils using chemical weathering index. Second Int Conf on Geomechanics in Tropical Soils, Singapore 1988;1:55 ± 61. Sun S. Chemical composition and origin of the Earth's primitive mantle. Geochim Cosmochim Acta 1982;115:15 ± 28. Tapia J, Tosiani T, Loubet M. Rare earth elements (REE) behaviour during

gabbro weathering in tropical rain forest of South Venezuela. Goldschmidt Conference, Toulouse. Min Mag 1998;62A:1493 ± 4. Van Der Weijden CH, Van Der Weijden RD. Mobility of major, minor and some redox-sensitive trace elements and rare-earth elements during weathering of four granitoids in central Portugal. Chem Geol 1995; 125:149 ± 67.