HAL Id: hal-00929664https://hal.archives-ouvertes.fr/hal-00929664

Submitted on 1 Jan 1999

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Kishk - a dried fermented milk / cereal mixture. 4.Microbiological quality

Adnan Y. Tamime, David Mcnulty

To cite this version:Adnan Y. Tamime, David Mcnulty. Kishk - a dried fermented milk / cereal mixture. 4. Microbiologicalquality. Le Lait, INRA Editions, 1999, 79 (4), pp.449-456. �hal-00929664�

Lait (1999) 79, 449-456© InraJElsevier, Paris

449

Original article

Kishk - a dried fermented milk/cereal mixture.4. Microbiological quality

Adnan Y. Tamime-", David Mcblulty''

a Food Standards & Product Technology Department, SAC - Auchincruive,Ayr KA6 5HW, Scotland, United Kingdom

b Biomathematics & Statistics Scotland, University of Edinburgh, James Clerk Maxwell Building,The King's Building, Edinburgh EH9 3JZ, Scotland, United Kingdom

(Received 6 March 1998; accepted 24 November 1998)

Abstract - The microflora of 25 samples of Kishk were determined. Pathogens such as Staphylo-coccus aureus, Escherichia coli, Salmonella sp., Listeria sp., Campylobacter sp., Yersinia sp. and Bru-cella sp. were not recovered from any of the samples at the levels tested. Enterotoxins of staphylo-cocci, bacilli and clostridia also were not detected. Bacillus cere us, aerobic spores, total viable cou ntand contaminants were recovered in appreciable numbers, and sorne B. cereus strains that may pro-duce toxins could he the major bacteriological risk associated with this product. Coliforms, faecal ente-rococci and Clostridium perfringens were only recovered from samples 7 and 1, respectively. How-ever, yeasts and moulds were recovered intermittently from sorne of the Kishk samples, and thepossible high cou nt of yeast (8.5 x 105 cfu-g ") could be attributed to the possible use of 'artisan' -typestarter culture containing lactose fermenting yeast. The lactic acid bacteria counts were up to3.0 X 103 cfu-g :' and 1.2 x 106 cfu·g-l for Jactococci and lactobacilli, respectively. The overall inher-ent characteristics of Kishk (0.4 water activity [Aw], 3.8 pH, high in salt and contains a wide rangeof organic acids) contributes towards the microbiological safety of the product. © InralElsevier,Paris.

Kishk / microbial count and toxin / water activity

Résumé - Le kishk - un mélange lait/céréales fermenté et séché. 4. Qualité microbiologique.La microflore de 25 échantillons de kishk a été déterminée. Les bactéries pathogènes telles que Sta-phylococcus aureus, Escherichia coli, Salmonella sp., Listeria sp., Campylobacter sp., Yersinia sp.et Brucella sp. n'étaient pas présentes dans les échantillons aux niveaux testés. Les entérotoxines destaphylocoques, de bacilles ou de clostridies n'ont pas non plus été détectées. Des quantités appré-ciales de Bacillus cereus, spores aérobies, flore totale et contaminants étaient dénombrées; cer-taines souches de B. cereus qui peuvent produire des toxines constitueraient le principal risque bac-tériologique associé à ce produit. Des bactéries coliformes, entérocoques fécaux et Clostridium

* Correspondence and reprints. a.tamime@au.sac.ac.uk

450 A.Y. Tamime, O. McNulty

perfringens ont été trouvés seulement dans les échantillons numéros 1 et 7. Cependant, des levureset des moisissures étaient retrouvées par intermittence dans quelques échantillons de kishk, et lapopulation élevée de levures (8,5 x 105 ufc·g-I) pourrait être attribuée à l'utilisation d'un levain« artisanal » contenant des levures fermentant le lactose. Les dénombrements des bactéries lactiquesatteignaient 3,0 x 103 ufc·g-I et 1,2 x 106 ufc·g-1 respectivement pour les lactocoques et les lactoba-cilles. Les caractéristiques physico-chimiques du kishk (Aw de 0,4 ; pH de 3,8 ; forte teneur en sel;forte variété d'acides organiques) contribuent à assurer la sécurité microbiologique de ce produit.© Inra/Elsevier, Paris.

kishk / dénombrement microbien / toxine microbienne / activité de l'eau

1. INTRODUCTION

The traditionaI method for the manufac-ture of Kishk is very basic [24], and themicrobiological quality of the product ismainly govemed by factors such as: 1) themethod used to ferment the milk; 2) thehygienic conditions practised during themanufacturing stages including whether ornot the milk was heat treated (e.g. boiled)before the fermentation stage; and 3) thedrying stage in the open air in the sun. Threepossible types of lactic fermentation maybe distinguished during the manufacture ofKishk: first, unknown composition of thestarter culture where the indigenousmicrof!ora is used to ferment the unheatedmilk; second, 'artisan' starter culture is nor-mally used frequently by seeding the milkwith yoghurt or another type of fermentedmilk from the previous day's batch, andhence, the starter composition may be vari-able; and third, selected lactic starter cul-tures obtained from commercial sources ofknown composition.

It is evident that acidification of miIk canex tend the shelf-life of the manufacturedproducts, and limited data are avaiIable onthe microbiologicaI quality of Kishk. Nev-ertheless, the low moisture content « 10 %),acidic nature of the product (-3.8 pH) andthe addition of salt during manufacture(-2.8 g·100 g-l NaCI in the dried product)may suggest the microbiological safety ofKishk [25].

Faecal enterococci count at a level of3.4 x 102 colony forming units (cfuj-g! wasreported by Atia and Khattab [2] in onlyone of eight Egyptian Kishk samples tested.However, the groups of undesirable microor-ganisms, which were found in different com-mercial sampIes of Kishk, were mainlyspore-formers (i.e. Bacillus spp.), yeasts andmoulds [24]. In Iran, the death of two peo-ple, who had clinical symptoms of botulismfood poisoning, was associated with the con-sumption of Kishk, and Haydarynia [9]reported the growth, survivaI and produc-tion of toxin of Clostridium botulinum inlaboratory-made Kishk.

In previous studies the chemicaI andnutritional properties of Kishk were detailed[25-27]. The objective of this study was toinvestigate the microbiological evaluationof 25 different samples of Lebanese Kishk.

2. MA TERIALS AND METHODS

2.1. Kishk sampIes

Twenty-five samples of commercial Kishk(-1 kg each) were obtained from different retai!outlets in Lebanon [25].

2.2. Analytical methods

2.2.1. Water activity (A) and pH

The water activity in the Kishk samples wasdetermined according to the method described

Kishk - microbiological quality

by O'Brian [21]. Ail measurements were madewith an Aw Portimeter Dewpoint Meter DP 383R(Portimeter Ltd., Marlow, UK) placed in a warmroom at a nominal temperature of 25 oc. As theKishk products had low Aw' equilibration of thechamber for 24 h was employed. A standard solu-tion of NaCI was checked weekly as a control.

The pH in the Kishk sampI es was measuredaccording to the method described by Tamime etal. [25].

2.2.2. Microbiological analysis

The Kishk samples were transported to Scot-land by air at ambient temperature and stored at-5 "C on arrivai for around 6 months. A widerange of microbiological tests including toxindeterminations were carried out, and table 1 sum-marises such analysis, including the methodsused. The preparation of each sample and dilu-tions for microbiological examination was carriedout according to the method described by lOF[20]. Each Kishk sample (10 g) was homogenisedin 90 mL of sterile 2 % sodium citrate solution at-7 "C for 15 s in a Colworth Stomacher400 model BA 6021 (A.J. Seward Medical, Lon-don, UK). Seriai dilutions were made in / 1/4strength sterile Ringer's solution and plated induplicate on specifie agar, and the average of thedeterminations was expressed as cfu-g ".

2.3. Statistical analysis

The data were analysed using graphicalexploratory data analysis techniques availablein Minitab 10.2 (copyright Minitab,lnc., 1994).

3. RESUL TS AND DISCUSSION

3.1. Chemical and nutritionalcomposition of Kishk sampi es

The average chemical composition (pro-tein, fat, carbohydrates, ash, organic acidsand fatty acids) and micronutrients (min-erai, selenium, amino acids and vitamins)including ~-glucan, fibre, phytic acids con-tents and profiling of the proteins of 25 sam-pIes of Lebanese Kishk have been reportedby Tamime et al. [25-27].

451

Table I. Methods used for the microbiologicalanalysis of Kishka.

Tableau I. Méthodes utilisées pour l'analysemicrobiologique du kishk".

Microorganism/test method

Reference

Total viable count''Total microbial contaminants''Psychrotrophic countColiforms (MPN)dStaphylococcus aureusLactic acid bacteria"Yeasts and moulds!Listeria monocytogenesSalmonella spp.Escherichia coli (MPN)Bacillus cereusClostridium perfringensEnterococcus faecalis (MPN)Aeromonas hydrophilaPseudomonas spp.Campylobacter spp.Clostridia (MPN)Aerobic spore former countYersinia spp.Brucella spp.

[17][16][18][II][14][12,19][15][13][10]]]]]g [4]]]][7][8][22,23][6]

aMost of the tests used in this study were performed onspiral plate technique as opposed to spread or pourplate; b total count of bacteria, yeasts and mou Ids;C non-lactic acid bacteria, yeasts and moulds; d MPN:most probable number; C thermophilic lactic acid bac-teria mainly to enumerate yoghurt starter cultures;f oxytetracycline was replaced by chloramphenicol;g test methods were described by the Oxoid Manualincluding the enterotoxin detection of S. aureus,B. cereus and C. perfringens.

a La plupart des tests dans cette étude ont été réaliséspar ensemencement spiral plutôt que par étalement ouensemencement dans la masse. b Dénombrement totaldes bactéries. levures et moisissures. C Bactéries non-lactiques, levures et moisissures. d MPN - nombre leplus probable. C Bactéries lactiques thermophiles (prin-cipalement pour dénombrer les levains du yaourt).f L'oxytetracycline a été remplacée par du chloram-phenicol. g Les méthodes sont décrites dans le manuelOxoid avec la détection des enterotoxines de S. aureus,B. cereus et C. perfringens.

452 A.Y. Tamime, D. McNulty

3.2. Water activity (Aw) and pH

The residual microbial activity in foodproducts is dependent on a multitude of fac-tors such as availability of nutrients, stor-age conditions, acidity, salt level and/orwater activity. The Aw readings of25 sam-pies of the Kishk ranged between 0.34 and0.43 (table II), which are proving to berather low to support the growth of microor-ganisms that not even the most xerophilicmou Id cou Id grow.

The salt, pH and certain organic acidscontents of these Kishk samples have beenalso included in table Il for comparativepurposes where such data may provide therelevant information that may influence thesurvival of certain microorganisms in thedried product. For example, the pK, of lac-tic < acetic < propionic acids [5], and theundissociated acid is more active as anantimicrobial to inhibit the growth ofpathogens and other undesirable microor-ganisms [1].

3.3. Microbiological quality

The microorganisms that have beendetected in the Kishk samples are shown intable III. The total viable count (TVC)

in 20 samples of Kishk ranged between3.1 x 102 and 1.1 x 106 cfu-g ", No TVCwas recovered from Kishk samples 21, 23and 24, suggesting that the fermentedmilk/cereal mixture was subjected to heattreatment before drying; however, counts forsamples 4 and 5 could not be carried out dueto the presence of spreading colonies, prob-ably Bacillus sp. Similar pattern of countswere observed when the total contaminantcount (i.e. 2.8 x IOZ to 1.3 x 106 cfu-g") wasperformed (table II/), including Kishk sam-pies 4,5,21,23 and 24 (for sample identi-fication, refer to [25]). Similar total countsof commercial and laboratory-made Kishkhave been reported elsewhere [24]. How-ever, the low microbial contaminants countranged between < 10 and 4.7 x 102 cfug!for Kishk samples 1, 2, 7, 9, 10, 21, 23 and24, whilst highest counts (1.3 x 105 to1.3 X 106) were recovered from samples 3,12, 13, 16 and 17. The high microbial con-tamination in these Kishk samples reflectthe poor sanitary conditions during the man-ufacturing stages or post-production con-tamination, and interestingly four of theseproducts were made by non-dairy organi-sations.

The presence of coliforms were onlydetected in Kishk samples 3 and 13 whichyielded 24 and 43 MPN (most probable num-

Table II. Measurements of available water, pH, salt and certain organic acid contents of different sam-pies of Lebanese Kishk.Tableau II. Détermination de l'eau disponible, du pH, du sel et de certains acides organiques dansdifférents échantillons de kishk libanais.

Component Minimum Maximum Mean

Water activity (Aw) 0.34 0.43 0.40Salt (g- 100 g-l)' 1.0 4.5 2.9pH' 3.63 4.12 3.77

Organic acids (mg·g-1)'

Lactic 18.75 43.87 32.47Acetic 0.40 0.86 0.59Propionic 0.92 7.45 3.55

a Data compiled from [25].a Données issues de Tamime et al. [25].

...<)

.nE;:l

=<)

::ëCIl.n20..

Kishk

-m

icrobioJogicaJquality

M~

~~~

~MM

MMMMMM

MMMM

dd~dd~d~dd~d~dddddd~d~ddd

vv~vv~vvv

vvvvvvv~v

vvv

MM

MMMMMMMMM

MMMMMMMMMMMM

dd~dddddddddMdddddddddddd

Vv

VV

VV

VV

VV

V~V

VV

VV

VV

VV

VV

V

NN

'"

000X

XX

~r-000.,tN"";

~N

~N

NM

~~

~~

~~

MM

MM

MSSSSSSSSSSSSSSSoSooS

~XXXXXXXXXXXXXXX~X~~XOOr-OOO~~O~~OMM~O~

0~

N~OON~.,t"";"";N~"";"";N~"";~

~

453

§:ê'0uonëCIl

=.~ëou

1..~

<)

o...nE

ECIl

;:lC/.l=

454 A.Y. Tamime, D. McNulty

ber), respectively. Faecal enterococci wereonly recovered from Kishk samples 3, 6, 8,Il, 13,20 and 22 at low levels (table Il!). Itmay be significant to point out that samples3 and 20 had relatively high pH (4.1) whencompared with the rest of the Kishk sam-pies (i.e. pH :::;3.8), and both samples hadcoliforms and faecal enterococci. The lowcounts of these indicator organisms in Kishkcou Id be attributed to the die-off as a resultof low ~ and low pH of the product (table Il).In addition, the psychrotrophic count of1.0 x 103 cfu-g' was only obtained in Kishksample 3.

Aerobic spores were present in 17 sam-pIes of Kishk in numbers ranging from6.1 x 102 to 1.43 X 106 cfu-g! (table Il!).However, seven of the Kishk samplesyielded B. cereus at levels ranging between6.0 x 102 and 1.34 x 105 cfu-g:", Similarcounts have been reported by Atia andKhattab [2] on Egyptian Kishk. The pres-ence of B. cereus in Kishk is not surprisingas this microorganism is one of the com-mon contaminants of milk, and is mostlikely to be present in the parboiled crackedcereal (Burghol) (Khaskheli and Tamime,unpublished data). B. cereus produce sporeswhich are likely to survive in the Kishk.Sorne strains of B. cereus produce toxinswhich may cause food poisoning, and itcould be considered the major bacteriolog-ical risk associated with this product; how-ever, B. cereus toxin was not detected inany of the Kishk samples. Furthermore,C. perfringens was only recovered fromKishk sam pIe 3 at a very low level (3.0 x10-1 cfu-g ") which may not be a significantcount bacteriologically, and no toxin wasdetected. The MPN of clos tridi a were 2.3and 0.92 which were recovered from Kishksamples 3 and 7, respectively.

The yeasts and mould counts were recov-ered intermittently from the samples(table /II). Moulds were only detected inKishk samples 2, 9 and II, and the countswere6.0x 102,1.6x 102and 1.0x l02cfu·g-i,respectively. The yeasts counts rangedbetween 8.0 x 102 and 8.54 x 105 cfu-g-,

and were recovered only in samples 2, 3, 6,8, Il and 13. Slightly lower yeasts andmou Ids counts have been found in Kishk[24], but the high yeast count could beattributed to the 'artisan' starter culture usedwhich may contain lactose fermenting yeast[3].

With respect to lactococci and lacto-bacilli counts, spreading colonies appearedin the majority of the agar plates whichmade the LAB counts in the Kishk very dif-ficult, and hence, the efficacy of the finalcount may be inaccurate. However, sorneof the counts ranged between 1.4 x 102 and3.4 x 103 cfug :' for Lactococcus sp. and1.3 x 102 and 1.2 x 106 cfu-g! for Lacto-bacillus sp. (counts are now shown).

In this survey, Staphylococcus aureus,Escherichia coli, Aeromonas sp., Pseu-domonas sp. psychrotrophic count and Bru-cella abortus were not recovered from anyof the Kishk samples at the level tested (i.e.10-1 dilution), and Salmonella sp., Listeriasp., Campylobacter sp. and Yersinia sp. werenot recovered in 25 g of the product.

Figure / iIIustrates a matrix plot for TVC,contaminants count and aerobic sporescount. When the labelled outlier Kishk sam-pies were omitted from the calculations, ailthree variable microbiological counts werestrongly correlated. The outlier samples didnot have anything in corn mon with eachother nor were they distinguished from theother samples except when plotted. Therewere insufficient colony counts for the enu-meration of the remaining microorganismsshown in table /11to examine these counts indetail. However, when the microbial countsin figure / were related to variables (i.e.source, type of milk or packaging system;see table / in [25]), there did not appear to beany correlation between them.

4. CONCLUSION

Given the conditions during the manu-facture of Kishk, the overall indicator

Kishk - microbiological quality

1 Total ViableCount

i

455

·17 ~

. .. Contaminants..... .-.1•• 0')

·17 . ~.. .. . Aerobic. . Spores. 3 • . 3·

•• 1 '6 ... ·6. . .·11 .11

Figure 1. Scatter plot of totalviable cou nt, contaminantscou nt and aerobic sporescount of Kishk samples.Figure 1. Représentation dela dispersion des dénombre-ments en flore totale, conta-minants et spores aérobiesdans les échantillons dekishk.

microorganisms are only present in lownumbers in the product and in microbio-logical terms, are relatively safe and stable.This could be attributed to a multitude offactors such as: low water activity, high saltand the presence of a wide range of organicacids and low pH. However, in present-dayterminology the se would be considered as'hurdles', and it is interesting to observehow a traditional method of manufactureeffectively mimics present-day food safetyprinciples at least as far as preventing thegrowth of spores and B. cereus toxin. Inaddition, the intended use of Kishk is rehy-drating with water followed directly bycooking to prevent the growth of undesir-able microorganisms. There are indicationsin these microbiological results that theproduct quality is variable, with sorne sam-pies showing noticeably more contamina-tion than others. This is probably a reflectionof the standards of hygiene applied duringproduction and the quality of raw materialsused, mainly the Burghol, and it indicatesthat with improved hygiene, the microbio-logical quality of the Kishk could beimproved.

ACKNOWLEDGEMENTS

The authors thank Mr. S. Crawford and Ms.L. Drysdale for ski lied technical assistance. SACand BioSS receive financial support from theScottish Office of Agriculture, Environment andFisheries Department (SOAEFD).

REFERENCES

(1] Adams M.R .• Hall C.l .. Growth inhibition offood-borne pathogens by laetie aeetie aeids andtheir mixtures. Int. J. Food Sei. Teehnol. 23(1988) 287-292.

(2] Atia I.A .• Khattab A.A., Mierobiologieal andehemieal studies on Kishk, Alexandria Sei.Exehange 6 (1985) 63-71.

(3] Baroudi A.A.G .• Collins E.B .• Mieroorganismsand eharaeteristies of Laban. l. Dairy Sei. 59(1976) 200-202.

(4] Bridson E.Y .• The Oxoid Manual, 6th edn .• Uni-path Ltd., Basingstoke, 1990. pp. 2: 32-33.39-41. 56-57, 70-71, 174-179 and 10: 1.

(5] Doores S., Organie aeids, in: Davidson r.M .•Branen A.L. (Eds.), Antimierobials in Foods,2nd edn .• Marcel Dekker, New York. 1993.pp. 95-136.

(6J Farrell 1.0 .•The development of new selectivemedium for the isolation of Brucella abortusfrom eontaminated sources. Res. Vet. Sei. 16(1974) 280-286.

d

456 A.Y. Tamime, D. McNulty

[7] Freame B., Fitzpatrick B.W.F., The use of dif- [18] IDF, Milk. Enumeration of psychrotrophicferential reinforced clostridial medium for the micro-organisms - colony cou nt technique atisolation and enumeration of clostridia from 6.5 "C, Standard 101A, International Dairy Fed-food, in: Shapton D.A, Board R.G. (Eds.), Iso- eration, Brussels, Belgium, 1991.lation of Anaerobes-5, Academie Press, Lon-

[19] IDF, Yogurt. Identification of characteristicdon,1971,pp.49-55.micro-organisms (Lactobacillus delbrueckii

[8] Harrigan W.R., McCance M.E., Laboratory subsp. bulgaricus and Streptococcus salivariusMethods in Food and Dairy Microbiology, subsp. thermophilus), Standard 146, Interna-revised edition, Academie Press, London, 1976, tional Dairy Federation, Brussels, Belgium,pp. 176-177. 1991.

[9] Haydarynia A., The survival of Clostridium[20] IDF, Milk and milk products - Preparation ofbotulisium in Kushk, 23rd Int. Dairy Congress

BriefCommun., Vo!. l, 1990, p. lOS. samples and dilutions for microbiological exam-

[10] lOF, Milk and milk products - Detection of ination, Standard 122B, International Dairy Fed-

Salmonella, Standard 93A, International Dairy eration, Brussels, Belgium, 1992.

Federation, Brussels, Belgium, 1985. [21] O'Brian F.E.M., The control ofhumidity by sat-[II] IDF, Milk and milk products. Enumeration of urated salt solutions, J. Sci. Instrum. 25 (1948)

coliforms - colony count technique and most 73-76.probable number technique at 30 "C, Standard [22] Schiemann D.A, Developments of a two-step73A, International Dairy Federation, Brussels, enrichment procedure for recovery of YersiniaBelgium, 1985. enterocolitica from food, App!. Environ. Micro-

[12] IDF, Yogurt. Enumeration of characteristic bio!. 43 (1982) 14-27.microorganisms - colony cou nt technique at

[23] Schiemann D.A., Comparison of enrichment37 "C, Standard 117A, International Dairy Fed-eration, Brussels, Belgium, 1988. and plating media for recovery of virulent strains

[13] IDF, Milk and milk products. Detection of Lis- of Yersinia enterocolitica from inoculated beef

teria monocytogenes, Standard 143, Interna- stew, J. Food Protect. 46 (1983) 975-964.

tional Dairy Federation, Brussels, Belgium, [24] Tamime A.Y., O'Connor T.P., Kishk - a dried1990. fermented rnilk/cereal mixture, Int. Dairy J. 5

[14] IDF, Milk and milk-based products. Enumera- (1995) 109-1280.tion of Staphylococcus aureus - colony count [25] Tamime AY., Barclay M.N.I., Amarowicz A,technique at 37 "C, Standard 145, InternationalDairy Federation, Brussels, Belgium, 1990. McNulty D., Kishk - a dried fermented

milk/cereal mixture. 1. Composition of gross[15] IDF, Milk and milk products. Enumeration of components, carbohydrates, organic acids and

yeasts and moulds - colony count technique at fatty acids, Lait 79 (1999) 317-330.25 -c, Standard 94B, International Dairy Fed-eration, Brussels, Belgium, 1990. [26] Tamime A.Y., Barclay M.N.I., Law AJ.R.,

[16] IDF, Butter, fermented milks and fresh cheese. Leaver G., Anifantakis E.M., O'Connor T.P.,

Enumeration of contaminating micro-organisms Krause L, Kishk - a dried fermented milk/cereal

- colony count technique at 30 "C, Standard mixture. 2. Assessment of a variety of protein

153, International Dairy Federation, Brussels, analytical techniques for determining adulter-

Belgium, 1991. ation and proteolysis, Lait 79 (1999) 331-339.

[17] IDF, Milk and milk products. Enumeration of [27] Tamime A.Y., Barclay M.N.I., McNulty D.,micro-organisms - colony cou nt technique at O'Conner T.P., Kishk - a dried fermented30 -c, Standard toos. International Dairy Fed- milk/cereal mixture. 3. Nutritional composition,eration, Brussels, Belgium, 1991. Lait 79 (1999) 435--448.