2 KNAPP AND MaLELLAN: THE ESTIMATION OF CACAO SHELL. BY ARTHUR W. KNAPP B.Bu. F.I.C. AND BASIL G. MoLELLAN F.I.C. (Read at the Meethzq November 6 1918.) THE estimation of caoao shell has received the attention of many able chemists but it is doubtful if all the factors influencing the different methods or the 1imi.tations of their accuracy have been taken into account. In view of the exceptional facilities at our disposal for the examination of authentic samples we think that some record of our joint experience may be of use. VARIATION IN QUANTITY AND CHARACTER OF BHElili ON BEANS. (a) Variation due to Botanic Variety,-The different botanic varieties have distinctly different properties and this is one of the causes of the varying characters of the shell from cacao of different geographical origin.In many places the different varieties may be found growing on the same plantation. Thus taking beans from cacao-trees growing side by side in Trinidad and preparing the cacao in exactly the same way (without olaying) we obtained 13.3 per cent. of shell in Trinidad criollo, and 15.1 per cent. of shell on the Trinidad calabacillo (a more hardy variety with a smaller bean). (b) Variation due to Condition and Preparation of the Bean.-Not only does the original cacao vary but the character of the shell varies wibh the ripeness of the cacao when gathered. Cacao from the same plantation and therefore presumably, a mixture of the same botanic varieties was gathered in different stages of ripeness. Raw Unolayed Trinidad Uaoao. I Shell per Uent.Unripe . . . . . Ripe . . . . . . Overripe . . . 4 . . . Germinated . . . . ~~ ~ ~~ Unfermented . . 4 . . . Sweepings (pasoilla) . . . 18.5 39.9 The shell on commercial cacao consiets of true shell plus pulp and dirt. Every oountry has a different method of preparing the oaoao for the market and this affeot THE ESTIMATION OF CACAO SHELL 3 the character of the shell. Thus in those oountries where the cacao is not fermented or washed to remove the fruity pulp with which the bean is surrounded Borne of the pulp remains dried on the shell. Hence the high figure 18-5 per cent. for unfer-mented caca,o given above. More or less oare is taken in separating the broken beans, and about 5 per cent. of sweepings are produced containing 33 per cent. of shell.These sweepings sometimes get mixed with the whole beans. Washing.-In some countries the shell is washed quite clean the pulp being entirely removed. Thus waahed Ceylon gives only 8 per cent. of shell and whilst the shell of normal Accra cacao amounts to 12.2 per cent. that from washed Accra, averages only 9.5 per cent. This gives us some idea how much of the so-called shell is true shell and how much is dried pulp and dirt. Chyilzg.-As shell is sometimes used as human food the presence of clay is of interest. In the proparation of cacao the juicy bean has to be dried and whilst it is drying its outside is sticky and is liable to pick up any dirt with which it comes in contact. However some of the finest cacao-e.g. certain varieties from Trinidad and Venezuela-are deliberately coated with a thin layer of clay or earth.In our opinion the ash of clean shell does not exceed 8 per cent. (we find that the ash of Ceylon cacao shell is only 5 per cent.) ; the ash of those which are practi-cally free from earth-Arriba 6 per cent. Grenada Jamaica and Samoa 7 per cent., Accra 8 per cent. ; the ash of those which are slightly earthed Maohah and Bahia 9 per cent. ; whilst Trinidad gives 15 per cent. Columbian 17 per cent. and Ctlrupttno 24 per cent. From these figures it is evident that Trinidad cacao shell contains 7 per cent. of clay whilst Colurnbian contains 9 per cent. and Carupano 16 per cent. of red earth. We think one would hesitatie to recommend these last three shells for human food. The amount of clay is liable to considerable variation.Thus analyses of ten lots of Trinidad cacao gave-Shell . . . 13.5 to 17.4 per cent, Clay on bean . . . 0.3 ,) 2.7 ,) Clay on shell . . . 2.4 , 15.4 ,, Ash in shell . . . 10.4 ,) 23.4 ,, (c) Vmktion with CouNtry of Origin.-We give some figures below illustrating this but it will be reoognised that cacao from one country or even. from one plenta-tion is liable to vary considerably so that to give reliable average figures for any particular producing area would require many andyses. The average of six samples taken monthly in 1918 gave the following shell perctentages : Ceylon . . . . . Para . . . . . Bahia . . . . . Accra . . . . . Machala . . . . Arriba . . . . . Grenada . . . . Trinidad . . . . 7.9 10.4 12-1 12.2 12.5 12.7 14.5 15.8 These figures are for raw shell ; they were obtained by separating the shell from the beans by hand.Roasting reduoes the appaxent percentage of shell by 1 to 3 per cent 4 KNAPP AND MoLELLAN: Ether (0'720) Extract (per Cent.). I Size of Particles of Shell (Inches). We now propose to deal critically with the various processes which have been suggested for the estimation of shell. Fibre on Fat-Free (per Cent.). 1. CRUDE FIBRE. (a) O$icial Method of the Association of Omial Agricultural Chemists U.S.A. also Organic Analysis," 1909 vol. i. p. 70.) called Weender's and also Henneberg's Process. (See Allen's u Commercial Results obtained by this method : CBUDE FIBRE PERCJENTAGE ON FAT-FREE MATERIAL. ~~ Roasted Nib. I Roasted Shell, Liihrig (Zeitsch.Untersuck. Nahramq. #erwsm., 1905 14 263-267 ; see J. SOC. Chem. l d . 1905, Booth Cribb and Richards"'(ANA;]-I.BT i909 3; Winton Silverman and Bailey (Connecticut Expl. 24 341) . 134) . . . . . . . Station Report 1902) . . . . . . -4.7 to 6.2 4-7 , 6.6 11.1 to 19.8 13.2 , 16.3 13.7 , 20.7 I -Efect of Size of Particles.-The above process applied to roasted nib in various degrees of fineness all of which were comparable with the size of the ordinary cocoa powder of oommerce gave the following results Very finely ground 5.9 to 6.5 per cent ; finely ground 7.3 to 8.5 per cent. ; slightly aoarser 9.3 to 11.6 per cent. This has an important bearing on the use of the process and shows the necessity for obtaining the samples in the same state of division.Thevariation is probably due to the remarkable difficulty which is oxperiend in extracting the fat from coarse cacao particles the fat protecting the particles from the action of the weak acid and alkali. On the other hsnd in the case of shell we found to our surprise that the size had very little influence on the fibre figure. Taking a mixed shell we ground it to various degrees of fineness and obttt,iined the following results : Whole shell. 0.083 to 0.125 0.033 , 0.055 0.022 , 0.033 0.006 , 0.011 0.004 , 0006 Less than 0.004 0*011 , 0.022 2.46 2.64 2.82 2433 3.41 3.96 4-62 6*85 40-3 15-1 1401 14.1 14.0 14.5 14.0 14. THE ESTIMATION OF CACAO SHELL 5 -Pure coco8 (The rise in the ether extract with increasing fineness is of considerable interest, suggesting that unless the cells are actually broken the extraction is incomplete and the butyro-refractometer figure for the extraots at 40’ C.rose with fineness from 57.0 to 60.0 or the refraotive index from 1-4639 to 1.4659. It is evident that the extraot contained a fair percentage of resins with the cacao butter.) The method under review is a general method for grains seeds etc. and 8s it promised to be of some value for cocoa products a modification was adopted to meet the peculiarities of this particular material. As seven experimenters all followed this process we give the details in full. Possibly the most important modification is the use of a filter-paper in place of linen for filtration. This probably accounts for the resulte being higher than those on the official process.PER CENT. OF CRUDE FIBRE ON DRY FAT-FREE MATEBIAL. Fat (Per Cent.), .-25.5 10 per cent. shell I- -23.6 I 1 (e) 16.5 18.3 17.3, 17.1 (f) 19.0 18.8 19-9 (g) 16-6 17.0,. 16.8, 16*5,16*5,17-3, 18.8 -- I 50 per cent. shell 20 per cent. 21.5 shell 1 J5.1 100 per cent. shell 5.1 Fry’s Labora-tory. (a) 5-89 (a) 7.50 (a) 8-80 a) 12-76 a) 19.20 €towntree’s Laboratory. (b) 5.4 5.6 (c) 6.0 6.2 (d) 6.8 6.8 -:b) 7.3 7.5 :c> 7.7 7.9 ;a) 9.1 9.1 Lc) 8-8 8.8 ( b ) 13.9, 13.8 (c) 14.2, 14.0 -( b ) 18.6, 18.1 (c) 18.2, 18-5 I Cadbury’s Laboratory. 1 Mean. -I 5.50 6.09 6.78 7-39 7.83 (e) 5.6 5.7 4.6 4-9, (f 5.7 5.9 5-9 6.0 (91 5.3 5.6 5.8 5*6, 54 5.6 5.2 5-6 (e) 6.9 6.9 (f) 7.4 7.7 8.0 (g) 6.4 7-0 6-9 7.6, 7.2 7.3 7.2 13.81 14.08 18.35 18.33 ~ (e) 12.6 12.8 (g) 12.5 12.9 12.8, 12-8 12.8 Mean.5.29 5.87 5-60 6.90 7.70 7.09 8.09 8.21 12 *68 12.78 17-29 19.20 17.06 8-63 13.14 18.46 Process adopted for Fibre Estimatiort.-Extract a quantity of the substance so a8 to have at least 2 grrns. of dry fat-free material for the estimation of the crude fibre. The fat may be removed by any approved solvent such as ether petroleum ether etc. Transfer 2 grme. of dry fat-free matter to a 500 C.C 6 KNAPP AND McLELLAN: Present. flask and add 200 C.C. of sulphuric acid of strength 1.25 per cent. by weight. Connect the flask with an inverted condenser using a rubber stopper.Bring to the boil and maintain the boiling for thirty minutes. Filter through ordinary filter-paper wash with boiling water until the washings are no longer acid. Rinse the substance back into the flask with not more than 200 C.C. of water (use hot water if preferred) and add 2% grms. of sodium hydroxide (stick flake or other form provided it is pure and approximately 98 per cent. strength). Boil at once and continue the boiling for exactly thirty minutes in the same manner as directed above for the treatment with acid. Filter at once rapidly either (a) through double tared or counterpoised filter papers wash with boiling water until the washings are neutral, dry in steam-oven to constant weight weigh incinerate completely and weigh again ; or (b) through a single filter-paper.Wash with boiling water until the washings are neutral and then wash the fibre into a taxed dish and evaporate to constant weight. The fibre is then ignited in the dish and the ash determined. The loss in weight in either method is considered to be crude fibre. Errors due to the Procsss.-We decided to determine how fax analysts following the prescribed process would agreo. Hand-picked shell-free cocoa and oocoa-free shell were prepared on a large scale and these were mixed in certain proportions, and examined by seven analysts who were unaware of their composition. The results are expressed to only one place of decimal but the averages are calculated on the actual figures obtained. In these tests we had the advantage of the co-operation of Dr.Bywaters (of Messrs. Fry and Sons). An exBmination of these figures shows the fibre estimation-(1) does not give such good duplicates as might be expected; (2) is greatly influenced by the personal equation. If an analyst only makes one estimation he may get an experimental error of 6 to 8 per cent on the shell percentage. However we will assume that he takes the mean of at least two estimations and that he accepts 5.8 as the fibre figure for oocoa and 18.5 as the fibre figure for shell then from the above table we obtain the following results : Found by Analysts- Mean. 10 20 cf 1 (9) 11 *4 7.8 9.9 10.2 !?6 12.2 6 *6 (4 18.5 20.1 18.7 14.2 - 14.9 17.5 (4 (4 Several analysts found small quantities up to 5-7 per cent.of shell in the pure cocoa and as will be seen from the table found from 6.6 to 12.2 per cent. in a cocoa containing 10 per cent. It would appear that even when the process is carefully defined it is necessary for each analyst to determine his own figures for pure coco& and pure shell from the average of a large number of resulte. 46.5 53.6 55 *4 46.0 - 46.7 50 I 49. THE ESTIMATION OF CACAO SHELL 7 We have made a few experiments with a view to increasing the sgreemenb between different analysts. (I) FiZtering.-We prefer filter-paper to linen because of the extreme finenese of the cocoa particles and also because of the difficulty of defining the exact mesh for linen. We have used 12.5 cm. Swedish and Whatman No. 1 and find they give very similar results.The objection to filter-paper is the difficulty of removing the material without bringing away Borne of the filter-paper itself. It may be that copper gauze with 100 meshes to the linear inch (Fairley and Burrell J. SOC. C?zem. Id, 1918 37 155~) or an even closer mesh would prove the most satisfactory filtering medium. (2) Strength of Acid and Alkali-We find that within narrow limits the variation in the strength of sulphuric acid and sodium hydroxide used is of little importance. (3) The Time of Boiling.-The time of boiling should be rigidly adhered to, especially in the case of the alkaline treatment. The rate of boiling might wikh advantage be standardised possibly by the use of an oil-bath at; 120° C. (4) I t will be found more convenient to wash off the acid residue with 175 C.C.of hot water and add 25 C.C. of 10 per cent. sodium hydroxide instead of using 2-5 grms. of the solid 8s given in the above process. Diferences due to Variation in the Nature of the Shell.-The percentage of fibre in the shell is naturally affected by the ripeness of the o&cao when gathered and the amount of fermentation it undergoes. Cacao from the game plantation in Trinidad was carefully collected in bulk in an unripe ripe over-ripe and germinated condition, and fermented separately Some ripe cacao was dried without fermentation. The mean of two estimations for fibre gave the following figures : Per Cent. Unripe . . . . . 14.3 Ripe . . . . . 17.0 0 ver - ripe . . . . 16.0 Germinated . . . . 21.2 Unfermented .. . . 19.0 These figures help one to understand the variations which occur with beans from different countries. Country of Origin.-The various beans were given a low-temperature roast (the interior of bean probably not exceeding 120° C. to 130' C.). The shell was picked off by hand (the germs being excluded) ground and passed through a sieve (diameter of holes 0.02 inch). It was then extracted for twenty-four hours and dried in a water-oven. Two analysts (a) and (b) determined the fibre by the method described, and each analyst did at least two estimations on each sample. The results are only expressed to one decimal place as to give more would be indicating that the process is oapable of greater accuracy than it is. The number of factors influencing the percentage of fibre is so great that an examinationLof the above table reveals no obvious relation between the fibre and the botanic variety of the bean its fermented condition or other natural property.If we asmme as before for caloulation purposes that shell contains 18.5 per cent. of fibre then if we have 20 per cent. of shell in the fat-free material we shall fin 8 KNAPP AND McLELLAN: 15'4 per cent. or 24.0 per cent. of shell according as the shell is that of Grenada or Accra beans. I t should not be forgotten however that as a rule commercial cocoa is a blend of two or more different kinds of bean. Our figures me in general agree-ment with those of Winton Silverman and Bailey who obtained a minimum of 13.7 a mazimum of 20.7 and a mean of 18.0. FIBBE IN ROASTED CACAO SHELL Cacao Shell.1 Fibre on Fat-Free. Arriba (low roast) . Greneda . . Machala . . Arriba (high roast) . Trinidad . . Accra (well fermented) Bahia . . . Para . . . Accra (common) . Ceylon . . . . . . . . . . . . . . . . . . . . . . Average . . . (4 14.9 15.3 16.5 17.8 20-4 19.0 20 -8 20.4 21-3 20.0 18.64 (b) 15.9 15.5 16.9 18.6 17 0 20.4 20.8 20.2 21.0 21.6 18.79 Mean. 15.4 15.4 16.7 18.2 18.7 19.7 2004 205 207 21.4 18.7 Efect of Roasting.-It is unlikely that the analyst will encounter raw shell in cocoa; however a small amount of raw shell is separated in cleaning the beans and this raw shell might conceivably be used in cocoa but the shell on the market is always more or less roasted.The Rurmise that roasting would increase the per-oentage of fibre was found to be correot. The figures given are the mean of two or three analysee on the same sample. With Ceylon (criollo) cacao shell we obtain raw 17.5 and roasted 20.4 (Winton Silverman and Bailey working on Caracas beans, found in the shell fibre per cent. raw 14-7 ; under-roasted 15.4; medium roasted, 16.5 ; over-roasted 16.6.) With West African (forastero) cacao shell we obtained raw 18-2 and roabted 20.6. With Arriba cacao shell low roast 15.4 and high roast 18.2. As the cacao bean is subjected to varying degrees of roasting according to the kind of cocoa or chocolate required all the analytical figures for shell are liable to variation from this cause. (b) KOT&J's Method (ANALYST 1898 23 47; and Zipperer Die Chocokcde-B'abrika-tion, 280 German Edition 1913).In this method 3 grms. of the defatted substance are treated with 200 C.C. of glycerol (1.23 sp. gr.) containing 20 grms. per litre of concentrated sulphuric acid, under the pressure of 3 atmospheres for one hour. I t is then filtrated whilst hot through an asbestos filter and after being washed suocessively with hot water, alcohol and ether it is weighed then burnt; and the ash weighed. The difference between the two weighings expresses the amount of ash-free crude fibre THE ESTIMATION OF CACAO SHELL 9 Crithbns olt tht? Methad-The method has met with considerable adverse ariticism (Gury ANALYST 1912,37 447) and has been modified by Filsinger (Zeitsch Analyt.O h . 1900 223) and by Matthes and Muller (Zeitsch. Umterszlch. Nahrangt. Genmsm. 1906 15 159 ; also ANALYST 1906,31 159) ; but as the figures obtained varied from 5.5 to 9-7 per cent. for nib and from 13.1 to 1 8.2 for shell the process has nothing to recommend it as compared with the simpler method previously described. 2. NITROGEN. The Kjeldahl-Gunning process was used for the determination of nitrogen and the utmost care exercised in obtaining pure nib-free shell and pure shell-free nib the samples being hand-picked. The shell contained no germs. The nitrogen was deter-minedon two separate samples of shell drawn on different days. In the first sample the nitrogen was determined direct on the ahell in the second sample on the fat-free material. As we were unable to obtain petroleum ether trichlorethylene was used for extraction; this solvent dissolves out some of the theobromine.NITBOGEN IN ROASTED SHELL, Cacao 8hell. Fat (per Cent.). Ceylon . . . Para . . . Arriba . . Machala . . . Grenada . . . Accra (well fermented) . . . Trinidad . . . Bahia . . . Accra (common) . Average . . 2.00 2.30 4.24 6-55 2.36 2.45 4.57 2.64 2.02 3.23 Nitrogen (per Cent.). I Calculated on 1 Estimated on I Fat-Freeo Fat-Free. Estimated en Original Shell. (4 1.76 1-99 2.08 2-43 2 -75 2-76 2 -86 2.96 3.18 2-53 (4 1.80 2.05 2.1 7 2.60 2-81 2.83 3.00 3.04 3.25 ( b ) 1.95 1-98 2.11 2.51 2.69 2 -67 3.01 3.01 3 -23 2-57 I 2.61 We found that eamples taken from different bags of the same consignment often gave good duplicates and two lots of British West African cacao purchased in Accra and in Liverpool both gave 2.86 per cent.of nitrogen. We do not suggest that the figures in the table oan be taken 8s giving the average percentage of nitrogen in the shell on beans from a partiaular country. To give figures whioh couId be so applied many andyses would have to be made. The figures do however give some idea of the range of nitrogen figures for ehell. The table below serves the same purpose for nib. The nib oame out of the shell in which the nitrogen was determined above (oolumns marked a) 10 KNAPP AND MoLELLAN: 1.98 2-04 2.27 2.81 2.66 1.69 -NITROGEN IN ROASTED NIB. 2.08 4.75 5 -01 2 *36 4-90 5.29 2.38 5.31 5 *I3 3.18 - 3 -25 2 -86 - 3-00 1-76 - 1-80 I I Nitrogen (per Cent.).I- -Trinidad . . Bahia . . . Para . . . Accra (common) Ceylon . . . Accra (well fermented) Mach& . . Grenada . . Arriba . . . 53.94 56-14 55.03 55.47 52.08 52 -76 52 -38 53.96 51.73 -I 2 @49 2 -32 2-77 1.74 1.54 2.61 2.19 2 -07 2-25 -I 2-07 2.08 2-19 2-22 2.38 2-36 2 *39 2.38 2-73 4-75 5.01 5-19 5.19 5.13 5.29 5 *26 5.41 5.91 Average . . 1 53.72 1 2.22 I 2.31 I 5.23 E$ect of Rmstirtg.-This is of interest because of the very different &mount of roasting which beans undergo for different purposes. NITROGEN (PER CENT.) I On Original. 1 Calculated on Dry Fat-Free Nib.I - I I I 1 Raw. 1 Boasted. I Raw. 1 Roasted. - I I I I Bahia nib . . . Accra (well-fermented nib) . . . Ceylon nib . . . Bahia shell . . . Accra (well-fermented shell) Ceylon shell . .-' . The variation due to roasting is appreciable. Winton Silverman and Bailey found only slight changes but other experimenters (Zipperer and Ridenour) have found even greater differences. It is reasonable to suppose that in thoee cases where the cacao is moistened with alkaline carbonates before roasting some nitrogen would be lost. Criticism.-The value of this figure for the estimlltion of shell in cocoa is small. However it has the advantage of being a Egure which can be very accurately determined. The variation in the percentage present in fat-free nib is appreciable (we find from 4.75 to 5.91; Winton Silverman and Bailey found 4.74 to 5*41), whilst that in shell is very great (we find from 1-80 to 3.25; Winton Silverman and Bailey found 1.87 to 3.41).This great variation is not remarkable when we conside THE ESTIMATION OF CACAO SHELL 11 that we have so great a variety of shell and that we have theobromine and caffeine as well as a complex mixture of proteins present. In a generd way those beans which give a low nitrogen figure for the nib have a high peroentage of nitrogen in the shell. I n any calculation the germs appear as cocoa since the dry fat-free germ containe about 6 per cent. of nitrogen. The danger of relying on the percentage of nitrogen to estimate shell may be shown by taking an extreme case. Suppose we imagine Bahia shell mixed with fat-free Arriba nib then we could have 25 per cent.of shell present before the figure came below the average for nib or 43 per cent. of shell before the figure came below that for Trinidad nib. This is only by way of illustration. as in practice one has to deal not with one kind of bean but with a blend. 3. ELUTRIATION AND FLOTATION METHODS. (a) Filsinger and Druwe’s Process (Zeitsch. 5jt. chm. 1899,5 27 ibid. 1903 9 161 ; ANALYST 1903 28 216). Two grms. of the finely powdered fat-free substance are placed in a porcelain dish with 100 C.C. water and boiled (with continuous agitation) until the powder is thoroughly moistened and the froth disappears. The dish is allowed to stand for five minutes after which the upper part of the liquid is poured off.It is again filled up with cold water which is allowed to stand and then poured off. This process is continued until the supernatant liquor is quite clear. The liquid is then given a circular motion and allowed to stand until the solid material is completely settled, after which the liquid is drained off. This is repeated until the supernatant liquor no longer contains any floating particles. The sediment is washed into a Gooch crucible dried and weighed. The weight obtained is multiplied by the factor 1-43, obtainad by Drawe and corrects for soluble matter in the sherl. The figure so obtained directly represents the quantity of shell present. Criticisms of the Nethod-Ulrich (Arch. Pharm. 1911 249 537) obtained for roasted nib 0.63 to 0.73 and for roasted shell 57.85 to 85-17.As a result of his figures he suggests that the factor of 1.43 obtained by Drawe should be increased to 1-72. The fact that one observer has to use a factor of 1.43 while another recommends 1-72 shows that the mothod is only compamtive and subjeot to personal error. There can be no doubt that the results obtained with this method are largely dependent upon the sizes of the particles of shell and nib and also on the manu-facturing treatment which the cocoa has received. (b) Macara’s PTOC~SS (Bolton and Revis ((Fatty Foods,)’ p. 304; and Baker a d Hultort ANALYST 1918 43 199). An outline of the process is given from which we have worked out the following more detailed procedure : Fat is completely removed by extraction with petrol from lsome 10 to 20 grmg.of the cocoa the extraction being continued until the petrol ceases to show even traces of fat when evaporated on a watch-glass. Five grms. of the fat-free material axe weighed out and ground up with water in a mortar. The mortar is then filled with water and after allowing a minute or two to settle the cocoa still in suspension 12 IINAPP AND McLELLAN : Boasted Nib. ~-0.28 0.40 0.90 1-40 0.40 0.40 0.10 0.28 is poured into a cylinder. (The dimensions of the cylinders used were 41 to 43 mm. diameter by 280 mm. in height.) The residue still in the mortar is ground up with a further quantity of water and poured off in the same way into the cylinder u t i l all the material has thus been washed into the cylinder. The cylinder is then filled up with water thoroughly shaken and allowed to stand for fifteen minutes.The liquid is syphoned off bo within a short distance of the residue lying on the bottom. Ten such elutriations take place aad the liquid is allowed to stand after each for the following times 15 minutes ; 15 minutes; 10 minutes ; 7 minutes ; and then 6 periods of 5 minutes eeoh. By this time practically no matter should remain in suspension. The residues are then washed on to tared filter papers dried and weighed and the ash of the residues deducted from the weights so obtained. I t has been found that it is quite convenient to do six samples at a time and that the time occupied in filling up and shaking the cylinders is practically equal to the time occupied in eiphoning off the liquor.By this process we obtained the following results showing variation with kinds of cacao : Roasted Shell. 19.4 20.4 23.9 24-7 26.9 29.6 30.7 25.2 RESIDUE OBTAINED FROM 100 GRMS. OF FAT-FREE. Cacao. Bahia . . . . . Trinidad . . . . Grenada . . . . . Java . . . . . Ocumare . . . ' Acora . . . . . Arriba . . . . . Mean . . . . Criticisms of the Method.-This process being a modification of the Fileinger-Drawe process is subject to the same criticisms. With cocoas which were known to contain a large percentage of shell and which had been submitted to the (' Dutch" process we have obtained figures for sediment which would clasrllify these as free from shell. For the experiments below the details of the process more olosely resembled those described by Bolton and Revis in that 10 grms.of defatted material were taken the periods of standing were 15 15 10 5 5 5 5 and 5 minates and the residue was dried and weighed in a tared platinum dish. Revis and Bolton recommend 10 grms. to be def&tted and the shell to be estimated on this 10 grms. in which case the Soxhlet thimble would have to be scraped clean. I t is obviously better to run 6 fat estimation at the same time and take an accurately weighed portion of the fat-free substance. The fat-free residue is ground in a mortar. I t is praotically impossible to standardiee this grinding which unfortunately considerably affects the reeult. Thus a fat-free cacao mass when ground for five minutes gave 38.7 per cent. residue THE ESTIMATION OF CACAO SHELL 13 and when ground for 10 minutes 26.8 residue.We decided to grind the material for fifteen minutes. In Revis and Bolton’s amount they state that with shell 90 per cent. of the residue can be distinguished as spiral vessels and solerenchymatous tissue. In practice we find the pure shell powder does not give more than 40 per cent. of suoh material the remainder being masses of cells of indefinite structure. In some experiments performed under our direction by Mr. J. H. Hicklin M.A. the greatest care was taken to extract the very last tram of fat the conditions for adding the water to the cylinder were standsrdised (it was found an advantage to hold the cylinder slightly at an angle so that the water ran smoothly down and did not form bubbles) and the procedure in siphoning was carefully oonsidered.The siphon was turned up one-tenth the height of the water (Le. depth occupied by 50 0.c.) ; it was always lowered gradually and was lowered until the bend just touched the upper surface of the sediment. The times of resting ” were most rigidly adhered to. In spite however of the most elaborate preoautions duplicates did not agree well and the results for 5 10 15 20 25 30 50 and 100 per cent. shell when plotted did not lie on a straight line, Shell per cent. . . 0 5 10 15 20 25 50 100 Residue per cent. . 2.2 3.5 2.4 3.5 3.8 3-2 7.1 29.2 I t should be noted that the above experiments were performed on shell which we ouraelves ground very fine but which was not sieved. Macara (Zoc. cit.) mentions that the percentage of husk varies according to the degree of fineness to which the oocoa has been ground.The figures below show that the effect is so great as to render the process useless. A cacao shell powder was separated into parts of various sizes. Size of Shell. 0.004 to 0009 inch 0.003 , 0.004 Less than 0.009’;nch Residue. I Caloulated Shell. -I 34-0 per cent. 12.5 ?, 0.6 ?, 100 per cent. 34 9, 0 I? As 80 per cent. of the ground shell was leas than 0-003 inch in diameter the significance of its giving a figure lower than that of cocoa is evident. Thinking possibly that we were not working the process in a satisfactory manner we sent samples containing this shell to a well-known analyst who had faifh in the process. In a cocoa containing 13.5 per cent.of shell he found 3 per cent. by the Macara prooess and in a cocoa containing 27 per cent. of shell he found 0 per cent., although by other methods he was convinced that shell was present. (c) Flotation Method-Gos ke’s Process of Flotation in Calcium Chloride Solzction (ANALYST 1910 33 162). One grm. of the dry fat-free coooa is mixed with a stoppered tube with 20 0.0. of a calcium chloride solution prepared by dissolving 107.1 grms. of calcium chloride i 14 KNAPP AND McLELLAN: 100 0.0. of water (this solution should have 8 sp. gr. of 1.535 at 30°C.). The stopper of the tube is then removed the mixture is heated to boiling for two minutes and while still hot submitted to centrifugal action for six minutes. The liquid portion is how removed and the almost solid sediment is collected on a weighed filter washed with hot water until free from chloride dried at a 100°C.and weighed. From the weight of the sediment of husk obtained the excess quantity of the latter present in the original cocoa is calculated. For this purpose it may be taken that cacao husk itself yields 38.7 per cent. of sediment and the amount of fat in Ghe cocoa must also be taken into consideration The figure 38.7 was the highest result obtained on the examination of a number of samples of cacao husk the average result being 24-5 per cent. From the final result obtained a deduction of 6 per cent. is made this being the quantity of husk yielded by ordinary commercial Criticisms of the Method.-We find the process has many difficulties in practice.It is to be regretted that Goake did not find some suitable solution of sufficient density other than calcium chloride. A solution whioh sets to a d i d mass the moment it ie allowed to cool is inconvenient in use. It renders difficult what is essential to success the obtaining of a homogeneous mixture of cocoa and solution and further it tends to set during the centrifugal action. In the presence of these difficulties we have failed to obtain satisfactory results with this method on pure cocoa and shell, and the figures on mixtures suggest that either the cocoa floats the shell up or the shell drags the cocoa down. Filsiager and Botticher (J. Soc. Chm. Ind. 1910 29, 1129) and Duboxs and Lott (J. In&. Eq-. Chm. 1911,2 251)) and Schenk Schmidt, Gorbing (Zeitsoh.uft. Chem. 1912 18 2011 have failed to get satisfactory results by this method whioh in our opinion is unreliable. Prooess swggested by Ku2us7cy (Zeitsch. Untersuch. Nahrungs. Genussm. 1912 21, 654).-This process also depends on the difference in speaific gravity of the shell partides and nib prticIes and is open to the same criticism as the Goake process. So also is the process suggested by H. Grosse. (See J. SOC. Chem. Ifid. 1916, 36 137). 4. PENTOBAN. (Tollens a d Xrober's Process. L f. Ladw. 1900 357 ; and HeTtlter a d Skertchly, ANALYBT 1899,24,178-183.) From the quantity of weighed blaok precipitate of furfurol-phloroglucinol the quantity of furfurol can be caloulctted by division by 1-84 or more correctly by means of the figures given in Tollens' table (Konig Unters.Landw. u. Gewerbl. wicht Stoffe 225). From the quantity of furfurol we can obtain the amount of pentosm as follows Purfurol-O*0104 grm. x 1.88 = pentosan, Several other methods have been suggested for cdoulating this figure and are Calculation FormuZa.-Let a= weight of farf'urol-phloroglucide p = weight of COCO&S. &8 follows : pentosan required. (a) Dekhr.-p = [ (" x -) I - 0.0104 grm.] x 1-88. i 1-8 TEE ESTIBUTION OF CACAO SHELL 15 (b) T o Z ~ S (J. SOC. Chem. Id. 1907 26 987).-p =(a + 0-0052) x 0*8824. (c) Kriiber (Allen CommerciJ Organio Analysis," vol. i, p. 403).-(For (d) From weight of phloroglucide obtain pentosan from Wiley tables per cent. (8) Formula Adopted.-p = (a + 0*0052) x 0.89. Results obtained by this method : weights of precipitate from 0.03 to 0.3 grm.) p = (a +0.0052) x 0-8866.PENTOSAN PERCENTAGE ON THE FAT-FREE MATERIAL. WarDier (Eeitsch. Unttwsuch. Nahr-Hehner and Skertchley (ANALYST, Liihrig and Segin (Zeitsch. Uitersuii Nahrmgs. Gervussm. 1906,16,162) Prochnow (" Dissertation Braunsch-R. Adan (Seventh Internat. Congress App. Chem. 1909 Section VIIIc., Dekker ( U I r i c h ' ' A T c ~ ' ~ ~ a r ~ . 19ii; Welrnan (Zeitsch. opt. Chem. 19ii; Ulrich (Arch. Phrm.,'l911,249,55ij 217tgs. ~enwsm. 1899 a 892) . 1899 24 178-183) . weig," 1908) . . . 194) . 249 549) . . . 17 500) . Roasted Nib. --3.9 to 4.6 2.3 , 3.6 3.9 to 4.3 R8W Nib. 5.0 2.7 2.5 to 46 3.8 , 4.0 2.7 , 3.8 4.3 , 4.8 3.9 to 4.3 e Roasted Shell.---7.9 to 9.9 -9.0 to 9.6 -8.2 to 9.7 ItaW 8hell. 8.9 7-6 to 11.2 -7.6 , 10.5 8.2 to 9.6 7.5 , 8.5 8-2 , 9.8 Criticisms of the Method.-The paper by Adan (loc. cit.) created much interest in this process but it has not stood the test of time. The process is tedious and requires great manipulative skill to obtain satisfactory duplicates and there is no general agreement as to the correct formula to calculate the pentosan present. The ahief objeotion is the wide range of values; thus we obtained from 3.3 to 5.2 for various roltsted nibs and 7.7 to 10.0 for roasted shell. It will be noted that the highest figure for nib approaches the lowest figure for shell. One of us found difficulty in getting good duplicates with phloroglncinol and so we tried the use of Fehling solution to estimate the furfurol following the process reoommended by Eynon and Lane (ANALYST 1912 37 41-44) and using Erober's table to calculate the pentosans.The results were little better the disagreement apparently being due to the different amounts of reducing bodies produced. Further, we found that the mere addition of boiling water to aid in grinding the shell with subsequent evaporation of the water greatly reduced the percentage of pentosans obtained one shell giving as low a figure as 4.7. Jolles (Zeitsck. Analyt. Chem. 1906 196) prefers to titrate the furfurol using bisulphite. 16 KNAPP AND McLELLAN: RRW Nib. Raw Shell. I I Devin and Strunk (Ulrich Arch. Pharm. 1911,249,549) Prochnow ( 1 6 Dissertation Braunschweig,” 1908) 1-0 to 1.6 3.0 to 5.3 .1 - 1 6.3 ) 8.6 C&ticiSms of the Method.-The refinement suggested by Jiiger and Unger does not improve matters appreciably the figures which we obtained for roasted nib being 2-6 to 4.3 and 8.8 to 9.0 for roasted shell natural variations which are great enough to render the process of little or no value. Prochnow concludes from his results, noting the great variations between them and those of Dekker and Liihrig and Segin, that the pentosan determination is of no great value for the determination of shell in cocoa. Ulrich states that his results and those of other investigatom prove that this method for determination of shell in cocoa has no value when the shell content is below 30 per cent. 5. ‘‘ COCOA RED ” METHOD. Ulrich (ANALYST 1912 37 52) obtained the following percentages of ferrio chloride precipitate on the fat-free material Roasted nib 107 to 16.8; raw nib, 12.1 to 17-1 ; and shell 0.Our figures show a wider range for nib and from 1.3 to 4-8 for shell. FERRIC CHLORIDE PRECIPITATE PER CENT. ON FAT-FREE MATERIAL. Cacao. Arriba . ‘ . . . Accra . . . . . , unfermented (1) . . . (2) . . . B&ia . . . . Trinidad [i) . . . . 9 , . . . 9 9 (2) 0 . - , unfermented . . . Grenada . . . . . Ocumare . . . . . Java . . . . . Mixed shell pure commercial samples (1) 9 9 (2) 99 : (3) 9 9 Y Y . . . . . . . . . I . . . . Roasted Nib. 22.1 22.0 14.0 12.7 21.2 20.0 9.6 12.2 17.0 15.7 15.6 -- -Boasted Shell THE ESTIMATION OF CACAO SHELL 17 Ulrich only claims for his method that the presence of 10 per cent.and upwards of shell can generally be ascertained.” This modest conclusion is arrived at by the consideration of a mass of figures expressed to four places of decimals. Ulrich ie apparently of opinion that only I‘ COCOB red ” is estimated. Actually other tannin compounds are precipitated at the same time (note the figures on unfermented nib above). He quite incorrectly assumes that the cocoa red is entirely absent from the oacao shell. The oriollo cacao bean is often almost white and cocoa red isdeveloped during fermentation and consequently a highly fermented cocoa is rioher in cocoa, red than one whioh has been mildly fermented. During the fermentation tannin compounds spread from the beans to the shell.Zipperer (Zoc. cit.) gives the cacao-tannic acid soluble in 80 per cenb. alcohol in raw shell as varying from 3.80 to 9.15 per cent. Taking these facts into aocount we cannot recommend Ulrich‘s lengthy process; posaibly when we have a more definite knowledge of the tannin compounds present in cacao a useful method may be devised for the estimation of cocoa in shell powder but the variation in the figure €or nib will always be fatal to its U E ) ~ for the estimation of small quantities of shell in cocoa. 6. METHODS INVOLVINU USE OF THE MICROSCOPE. Before the grinding of cocoa essence became such a fine art it was possible by means of the microscope to identify with certainty the larger particles of shell and to form an approximate idea as to the quantity of shell present.With the advent of modern machinery more thorough grinding of the cocoa ie effected and the distinc-tion between particles of shell and particles of nib becomes more difficult. Processee have been suggested in which use is made of stains for colouring the particles of shell or nib. (( RzLthenkm Red ” Ilfethod.-We find the following the best method of using this : Mix the sample with 5 per cent. solution of lead acetate. Add a few drops of ruthenium red dissolved in lead acetate. This stains the mucilage cells a bright pink. About 10 per cent. of an average shell powder becomes stained but the results are rather irregular and we find it impossible to distinguish between cocoas containing 15 per cent. and 30 per cent.of shell. This process was the subject of much discus-sion in the famous legal case in 1910. Detection of Cocoa Husk” (J. Chm. Tech. 1916 p. 17.)-This process depends on the theory that shell absorbs a solution of ferric chloride whilst COCO~ does not. Ferrocyanide is finally added which etains the shell blue. We experimented with this method and found that many particles of shell showing obvions shell structure do not stain. There is some danger of gebting a precipitate of ferric ferrocyanide which may be taken for shell. We tried removing the fat but found that fat-free shell did not stain and that the size of pwticles makes no difference. The process is valueless as much a8 30 per cent. of shell escaping detection. The method of Wadcky and Wimmer (ANALYBT 1916 41 46) requires ultra-violet light and a special microscope.I t suggested to us the use of the light from a, E. B. Gerrams 18 KNAPP AND McLELLAN : mercury lamp. By this light the cacao particles appear bluish-green and the shell greenish-brown. The opacity of the particles interferes and the method has no quantitative value. Of the various methods we have tried we prefer that worked out at our suggestion by Mr. Hicklin : New Method Phloroglucinol and Iodine Staining.-PhIoroglucinol was used as a possible stain for isolating shell tissue. I t was found that while this did not give a stain indicating lignified tissue-i. e. 'I cherry-red "-yet it cleared the shell particle8 and heightened the colour somewhat. After treatment with phloroglucinol iodine walj added before putting on cover-slip.This stained the starch of the cacao and left the shell untouched. I n this way shell particles were clearly isolated. Retails of Process.-Put about 025 grm. of sample in a very small beaker and add 2 to 3 C.C. of phloroglucinol (phloroglucinol 1 grm. rectified spirit 20 c.c. distilled water 80 c.c.). Mix thoroughly and allow to stand for fifteen minutes. Pour off the bulk of the liquid and transfer a small quantity to a slide. Add a spot of pure hydrochloric acid and spot of & iodine. Mix thoroughly with a glass rod and put on a cover-slip. In this form the preparation will soon dry and be useless for examination. Hence if if be desirable to keep the slide more than a few hours the mount should be treated as follows After adding iodine and stirring allow the slide to stand to dry protected from dust.When dry add drop of glycerol stir well, and plwe on the cover slip, We consider this process an improvement on any of the methods we have tried, but it is liable to all the usud errors associated with microscopic methods. Apart from the fact that with h e l y ground GOCO~S it is impossible to distinguish with certainty between particles of shell and particles of nib we have no definite proof that the eolouring agents used exercise an absolutely selective action. There is also the great question of sampling and the personal variation which must occur in preparing the microscopic slides. Consequently the microscopic proceases for cocoa although they give useful information in experienced hands cannot be regarded as giving quantitative results.A mixture of 20 coooas gave on sieving-0.004 inoh to 0.009 inch . . . 20 per cent. 0.003 inch , 0.004 inch . . . 74 ,, Less than 0.003 inch . . . . 6 ,, Sieving separates and measures aggregates not the actual particles. Under the microscope the average size of the cocoa particles was 20,~ (O*OOOS inch) and of the shell used in our experiments 30p (0*0012 inch). 7. COLOUR OF ETHEREAL EXTRACT. 0. Keller (Arch. Pharm. 1917 255 405) founds his process on his observation that the ethereal extract of pure cacao kernels is colourless or faintly yellow whilst the extract of husks is distinctly brown. The colours are estimated by comparison with a solution of ferrio chloride containing 0-1 grm. iron in 100 C.C.The extract from pure powdered bean is matched by 2.4 C.C. and shell by at least 3.5 o.c. the average being 4.4 to 4.5. The author's figures show that the process is useless for estimating shell. We investigated the procetw and found that the colour was no THE ESTIMATION OF CACAO SHELL I9 seriously affected by the wetness of the ether nor by the treatment of the shell with alkali nor by contact with iron machinery. But the figure obtained is very appreciably affected by the fineness of the shell ; the smaller the particles the greater amount of colour is extracted (for variation in amount of extract with fineness see under Fibre Estimation). Thus with We& African cacao we obtained the following figures Raw shell coarse pieces 3.0; finely ground 12.0. Roasted shell coarse pieces 6.5 ; finely ground 10.0.Alkdised shell coarse pieces 4-0 ; roasted shell, finely ground in iron mortar 14.0. 8. ASH CONTENT AND ALKALINITY OF THE ASH. The use of these to estimate shell was proposed by Luhrig (Zeitsch. Unterszcch. Nuhmngs. Qsnussm. 1903,12,161 and ANALYST 1905,30,206). For a general discus-sion of the ash of cacao shell see Baker and Hulton (ANALYST 1918,43,193). The wide variation in the ash content of the shell of various hypes of cacao due to treatment on the plantation has already been dealt with and the effect of the processes at the factory on the ash content is considerable. We find this figure of practically no vdue for the estimation of shell although iti gives useful hints on the method of manufacture of the cocoa.I n the paper by Baker and Hulton the mean analyses of pure shells were given. It will be of interest to give analyses of actual factory outputs-that is of the shell which might be used for purposes of adulteration. Below we give analyses of the principal commercial shells sold on the English market is bulk. These are obtained from mixtures of many varieties of cacao and give one an accurate idea of the kind of shell actually avaihble in quantity. ASH OF CACAO SHELL (1918) PERCENTAGE COMPOSITION. Shell. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. Average Total Ash. 14.42 11 *60 10.14 9.96 9.69 8-87 8.74 8.34 7-83 7-80 7.78 7.69 7-32 Ash Insolubll in Water. -2.89 7.77 5.70 6.99 4.67 3.33 4.07 3.53 3.02 3.58 3.29 3.00 3.07 9.24 I 4-22 Ash Soluble in Water.11.63 3.83 4-44 2.97 5-02 5.54 4.67 4.81 4.81 4-22 4.49 4.69 4.25 5.01 -Alkalinity of Water-Soluble Ash (a5 IiaO). 5.69 1.93 1.66 1-34 2-55 212 2 *28 2.23 2.34 1.91 1-72 2-19 2 *02 -2.30 Chlorine (as NaC1). 0.22 0.37 0.10 0.19 0.33 0.12 0.44 0.22 0.34 0.12 0.06 0.15 0.29 0.22 -Fat. 8-8 3.0 4.2 12.7 9.0 11.9 5.8 5.2 3.6 5-0 9.3 3.9 3.0 6.5 Moisture. 10.5 10-0 8.2 2.7 9.1 3.0 3.8 9.0 8.7 10.6 3.2 8 *I 9.0 7 *3 The alkalinity of the water-Boluble ash was determined on the filtrate by fitration with standard wid 20 KNAPP AND McLELLAN : 9.SOLUBLE SILICA PROCESSES. This prooess suggested by Zipperer Matthes and Muller (Zeitsch. Untersuch. Nahrungs. Genussm. 1906,15,95) has been proved by Matthes and Rohdich (Zeitsck. Oft. C h m 1908,14 166) to be quite unreliable. They find that from the examination of twenty specimen8 of cacao beans the amount of soluble silica varies from 002 to 0.88 per cent. 10. IODINE VALUE OF THE FAT (Welman Zeztsch. oft. Chem. 1901 7 500). It has been suggested that the percentage of shell present in cocoa could be rtficertained from the iodine value of the fat. As the range of iodine values for butter from nib overlaps the range for shell butter this method is obviously useless, and becomes even ridiculous when we remember that nib contains 54 per cent. of butter and shell only 5 per cent.11. COLD WATER EXTRAUT. I t seemed psobable that the cold water extract of cocoa would be much less than that of cacao shell. Actually they are almost identical if expressed on the fat-free and this figure is useless for estimation of shell. GERMS. Raw oac&o beans uontain 0.7 per cent. of <‘germs” (these &re the hard rod-shaped radicles of the seed). As we have stated the examples of nib and shell of which we give analyses are germ-free. Germs are usually very completely separated from the nib. In this separation some machines produce a residual containing up to 12 per cent. of germs. Hence although the quantity of germs in cam0 beans is small, shell is sometimes seen on the market containing a fair percentage of germs. The analytical figures for germs are therefore of interest in aonnection with the estimation of shell.The figures which we have obtained for germ8 are similar to thoee published by Richards (ANALYST 1918 43,214) Moisture 5.1 to 5 5 ; ether extract 4.1 to 5.6 On fat-free dry material : Nitrogen . . . . 6.1 Fibre . . - . . 3.45 Ash . . . . . 7.5 to 8.2 Soluble ash . . 4.2 ,) 5.1 Alkalinity (a6 &O) ofsolubi; a8h . 1.6 It will be noted that the presence of germs in added shell if ignored would cause one to under-estimate the amount of shelI from either the nitrogen or fibre figure. CortcZ.usiort.-We have seen that the chief difficulties in the way of an acourate determination of cacao shell are-(1) The natural variation in oacao shell due to botanic variety and to the distinctive treatment which the bean receives in each country; (2) that cac&o shell is not a definite chemical substance and hence the estimation can only be of a purely empirical charauter.We have reviewed the various proceases for estimating shell and have oonolude THE ESTIMATION OF CACAO SHELL 21 that with all its faults the only one employed by itself which is capable of giving results of any value is the estimation of the crude fibre. It was reported in the ANALYST (1918 43 205) that the method of estimating shell in cocoa was then receiving official consideration. We hope that the publication of our work may be found of assistance in any research which is being conducted. We find that there is no process which will determine so low a percentage as 5 per cent.but we consider that if the analyst deducts 8 per cent. from the shell percentage as the mean of a number of analyses he finds by the fibre process he may safely repart that the cocoa contains at Ieast the remainder. In the Cocoa Powder Order 1918 the amount of shell in cocoa powder Was defined-Grade A ” to contain not more than 2 per cent. of shell and Grade ‘‘ B ” not more than 5 per cent of shell. It will be clear from the above that in our opinion no analytical method by itself or in conjunction with others will enable the analyst to distinguish between coco& containing 2 per cent. and cocoa containing 5 per cent. of shell. In conclusion we should like to thank Mr. B. W. J. Warren F.I.C. and Mias G. R. Woodhead B.Sc. for their assistance.MEWS. CADBURY’B LABORATORIES BOURNEVILLE. MESSRS. ROWNTREE’S LABORATORIES YORK. DISCUFNON. Mr. P. A. EmIs RICHARDS considered the authors were in rather a pessimistic vein in that they condemned all the recognised processes for the analysis of cocoa without suggesting an improved method. However he quite agreed wibh them in the statement that when cocoa nib and shell were extremely finely ground the levigation process was of very little use. He himself preferred to base his conclusions as to the composition of a sample on a general consideration of all the analytical figures taken in conjunotion with a full microscopical examination. Mr. CRIBB said that he found himself in entire agreement with the main conclu-sion to which the authors of the paper had arrived-via that of all the methods suggested for the determination of the proportion of shell the estimation of the fibre gave the nearest approximation to the truth ; but at the game time he never thought it safe to express an opinion on one factor alone and preferred to base his conclusions on as many other determinations as possible.Microscopio examination should of course never be omitted as by it alone oould the actual presence of shell be established. In former days when cocoa was less finely ground and dso more uniform a~ regards the fineness of grinding he had found that counting the particles of mucilage under the microscope after staining with ruthenium red afforded valuable quantita-tive indications especially in the cam of small proportions of shell.He found that, counting horizontally and vertically (near to the diameters) of a uniformly distributed and thin preparation under a 4 inch cover-glass that the number of particles in the case of shell was from 90 to 120 while in the case of good nib it rarely exceeded 4, and was often between 0 and 2. Of course great care had to be taken to get th 22 THE ESTIMATION OF CACAO SHELL preparations of uniform thickness. Now that the majority of cocoas are ground 80 much more finely the method had lost much of its value except that it gave a sure indication of the absence of shell when such was the case. In view of the difficulty of obtaining ruthenium red he had recently been employing methylene blue instead. I t stained the mucilage well but if the solution was too strong the starch and other structural elements were also colourea.Where larger proportions of shell were present he had hoped that the estimation of the cocoa red would prove umful a ~ l that substance was supposed to be entirely absent from the shell but unfortunately his experience again coincided with that of the aathors as he also had obtained figures as high as 4.5 per cent. in the shell alone. He inquired whether the wthora had any knowledge as to the mucilage content of the shell of unripe as distinguished from ripe beans. In reply Mr. Kmpp said The various figures are all expressed on the dry fat-free material unless otherwise sfated. The figures given on the cacao beans represent extremes that would be unlikely to occur with a manufactured GOCOB ahoe this is always made from a mixture of beans. We agree with Mr. Cribb that the miorosoopic examination has qualitative value but we do not think that the average miorosoopist can obtain quantitative results. With regard to the ruthenium red test this stains the mucilage which OOCU~F) not in but on the seed-coa,t. I do not know if it is greater in amount on unripe ” beans but would not be surprised to find it absent altogether from washed beans. The suggestion of sieving would be of use in certain cases if applied with discretion but it was always possible that the shell might be finer than the coco& or the cocoa finer than the shell although we believe that if they were ground together the shell usurtlly came out coarser than the cocoa