72 ELSDON AND BAGSHAWE: SHREWSBCRY AND KNAPP PROCESS THE SHREWSBURY AND KWAPP PROCESS FOR THE ESTIMATION OF COCONUT OIL. BY G. D. ELSDONj B.Sc., F.I.C., AND CYRIL R. BAGSHAWE. (Read cat the Meeting, February 7 , 1917.) SINOE the Shrewsbury and f i a p p process was first published (ANALYST, 1910, 35, 385) various papers have appeared in connection therewith. Ross, Race, and Maudsley (ibid., 1911,36,195) seem to find no difficulty with the process, although they suggest that the maximum figure for butter should be increased.Cribb and Richards ( i b d . , 1911,36,327) do not see that the process has any advantage over the Reichert-Polenske process, but they have nothing to say of any difficulties of manipulation. Revis and Bolton': (ibid., 1911, 36, 334) state that they have found it impossible to obtain concordant results with the same samples, and they suggest a modification which, they say, gives satisfactory results, and which was favourably commented upon by Macara at the ensuing discussion; but the figures which these authors give do not bear out their claims.This point will be discussed later. &an- field (ibid., 1911, 36, 446) speaks more or less favourably of the method, but con- siders the only advantage it has over the Polenske method to be one of speed.Shrewsbury and Knapp subsequently replied to these criticisms (ibid., 1912, 37, 3). * This modification, as stated by Bolton and Revis, was only intended to obviate some apparent sources of error in the original process. It was evidently not critically pursued by them, as they state that the results obtained by the Reichert-Meissl-Polenske process are quite satisfactory and afford all the evidence required.-EDITOR.FOR THE ESTIMATION OF COCONUT OIL 73 Preliminary experiments made by the authors on a large number of samples of butter, using the original process, as described by Shrewsbury and Knapp, gave results varying from 24.5 t o 37.9, some of these results, therefore, being considerably higher than those obtained by Ross, Race, and Maudsley, and by Cranfield.The samples, however, were not of known purity, though they were normal in all other respects. Duplicate results were, moreover, not particularly satisfactory, for whilst in some cases good agreements were obtained, in other cases duplicates differed by as much as 3.5 to 4.0 c.c., and in the case of a mixture of equal parts of coconut oil and butter fat the following three results were obtained:.76-6, 73-0, 70.1. The process meets such an obviously long-felt want, in that it makes use of the typical acids of coconut oil-lauric and myristic, which together form more than 60 per cent. of the oil (ibid., 1913,38, 8)-in place of those which occur in smaller quantities, and which go to form the Polenske figure, that it was considered worth while to continue the experiments further, using all possible precautions. The measurements of alcohol and water were all made with carefully graduated pipettes, and the final measurement; of the alcoholic solution of the fatty acids was made in a graduated flask.The results were, however, little better, and the process in its original form had to be abandoned. Following the suggestion of Shrews- bury and Knapp in their second paper (Zoc. cit.), the separation of the fatty acids from the alcohol was allowed to take place in a bath a t 37" C., but the results were no better. As it is evident that the main possibility of error lies in the different rates of separation of the insoluble acids from the alcoholic solution, endeavours were made to overcome this.For example, the length of time given to the separation was in- creased by varying amounts, up to two hours, but the results obtained, although somewhat better and considerably higher, were still far from being satisfactory. Finally, the process in this form had to be abandoned, as it was impossible to standardise the rate of rise of the globules of the fatty acids.The modification due to Revis and Bolton was next tried, the process being carried out exactly as described by them. At first the results obtained seemed to be more hopeful, but once again duplicates did not agree well. The reason for this was found to be that the results obtained are affected to a surprising extent by slight variations in the strength of the alcohol used, the differences being due, therefore, to the varying amounts of water retained by the filter-paper.Various methods of drying were tried, but the only one found to be really satisfactory was to leave the filter-paper exposed to the atmosphere for some hours-say overnight. This was found to be considerably more satisfactory than immediately drying the flask and filter-paper in the oven, although in cases where results are required quickly this method may be used.With this modification it was found possible to obtain quite concordant results--.duplicate experiments frequently agreeing to 0.1 C.C. It was not found necessary, however, to boil under a reflux condenser, as complete solution was obtained before the boiling-point of the alcohol was reached.The next point to be considered was the most suitable strength of alcohol to be used. Shrewsbury and Knapp suggested 60 per cent. by volume, and this figure Various modifications were attempted.74 ELSDON AND BAGSEAWE: SHREWSBURY AND KNAPP PROCESS is also adopted by Revis and Bolton, but neither quotes any experimental work to show that this is the best strength of alcohol to use.In order to discover the most suitable strength, experiments were made with a sample of coconut oil, with one of coconut-oil-free margarine, made by mixing equal quantities of lard and premier j u s and with a mixture of genuine butters. The modified Shrewsbury and Knapp figure was then obtained on these, using alcohols of varying strengths.The strength of the alcohol is expressed in all cases in terms of the specific gravity at 15.5" C., this, in the opinion of the authors, being the most satisfactory method, as it is free from all ambiguity. The alcohol was prepared from industrial methylated spirit 64 degrees O.P. The results of the experiments are given in the following table : TABLE VARIATION OF THE SHREWSBURY AND KNAPP FIGURE WITH THE STRENGTH OF ALCOHOL.Sp. Gr. of Alcohol a t 15.5" C.. . 1 *905 I -910 I ,915 1 -9175 1 -920 I -925 I I--------- I I I I Margarine . . 25.0 14.2 10.8 . * .. - 1 29.8 1 18.4 1 1::; 1 1::; I 5.7 . . 105.4 104.3 102.4 102.0 102.0 94.6 Obtained { zz::zt oil * . From these results it seemed evident that the greatest difference between the figures for coconut oil and margarine would be obtained by using alcohol of sp.gr. 0.920. Further work, however, made it clear that this difference was by no meant3 the only factor to take into consideration, for when the figures for mixtures of butter and margarine of varying strengths with coconut oil are plotted it is found that the resulting curve is not a, straight line, and, moreover, that its shape depends upon the strength of alcohol used; this is shown in the two curves.The figures obtained with alcohol of sp. gr. 0.910 are much more susceptible to small changes in the amount of coconut oil than those obtained with weaker alcohol, sp. gr. 0.920, but there are serious disadvantages in the use of this stronger alcohol.Thus, it is not possible to get such concordant results (the solutions are unstable, and are thus easily affected by slight changes in the conditions of the experiment), and also the figures for different samples of pure butter extend over a wider range; thus, for alcohol of sp. gr. 0.910 the range is about 12.0 (22-34), whilst for alcohol of sp. gr. 0.920 the range is only about 4.0 (9-13).Taking these various factors into account, it appears that the strength of alcohol most suitable is sp. gr. 0.920 a t 15.5" C., and this strength has therefore been adopted by the authors; figures are given for alcohols of other strengths, however, both for purposes of reference and to substantiate the suggestions put forward. With alcohol of this strength concordant figures can be obtained with mixtures of coco- nut oil and butter and coconut oil and margarine containing up to about 60 to 65 per cent.of coconut oil. With mixtures containing more coconut oil than this agreements are not so good, although this is made up to some extent by the greater differences in the results obtained by slight differences in the proportion of coconutFOR THE ESTIMATION OF COCONUT OIL 75 oil present.Where, however, accurate figures are required in the case of mixtures containing upwards of 70 .per cent. of coconut oil, it is better to use alcobol of sp. gr. 0.918. The figures obtained by the use of alcohol of both strengths are given in the tables. The somewhat elaborahe washing process adopted by Revis and Bolton has been used, but experiment has shown that it is really not necessary, and that once washing of the cake of fatty acids with 50 C.C.of cold water is all that is required. It is important that the filtration be carried out as nearly as possible at 15.5" C. For all the experiments described in this paper a jacketed funnel has been used through which water at 15.5" C. is allowed to circulate, but this would not be neces- sary in ordinary routine work, except where the laboratory temperature was far removed from 15.5" C.It has been found that more concordant results are obtained by cooling the alcoholic solution to about 14" C., and violently shaking before the final adjustment to 15.5" C. is made. The Process Used.-The process finally adopted is as follows: 5-0 grms. of the fat are weighed out into a 300 C.C.flat-bottomed flask (Reichert flask), and saponi- fied by heating with 15 C.O. of glycerol-soda over a naked flame until the liquid suddenly clears. (The glycerol-soda, is made by mixing together 700 C.C. of glycerol and 200 C.C. of a solution of caustic soda made by dissolving 1 pound of " pure by alcohol " caustic soda in 1 litre of water.) One hundred and forty-five C.C.of boil- ing water (measured hot) me then carefully added drop by drop t b avoid loss by spurting, and finally 10 C.C. of a 10 per cent. (by volume) solution of sulphuric acid, when the flask is corked and thoroughly shaken. The contents of the flask are cooled by immersing the flask in cold water until the fatty acids have set to a com- pact cake; the liquid is then filtered.Twenty C.C. of boiling water are added to the fatty acids which are retained in the flask, the acids melted, and again cooled. The liquid is poured through the same filter as before, and the cake of fatty acids broken by shaking violently against the sides of the flask, and then transferred t o the filter-paper by successive quantities of 20 C.C.and 10 C.C. of cold water. The filter-paper is allowed to drain and then removed from the funnel, it being supported in the rim of the filter-stand. The flask is inverted almost, but not quite, verti- cally over the filter-paper, and the whole is allowed to drain overnight. In the morning the flask is heated in the water-oven for about five minutes, and a current of air is blown through, the filfer-paper being then added and the drying continued for a further ten minutes.One hundred C.C. of alcohol (industrial methylated spirit), sp. gr. 0-920 at 15.5" C., are next added from 8 carefully graduated pipette, the flask corked, and the alcohol heated until the fatty acids have completely dissolved. The alcohol is then cooled below 15.5" C., thoroughly shaken, and allowed to stand in water at 15.5" C.for half an hour or until its temperatureis exactly 15.5" C. It is then filtered, and 50 C.C. of the filtrate titrated with & caustic soda after the addition of 1 C.C. of 0.2 per cent. phenolphthalein solution. In the case of fats containing large quantities of coconut oil the end-point is rather vague, and a little practice is necessary before it can be judged exactly. No water should be added.76 ELSDOX AND BAGSHAWE: 8inREWSBURY AND KNAPP PROCESS THE ESTIMATION OF COCONUT OIL IN MIXTURES.( a ) Mixtures of Oleo-Margarine and Coconut Oil.-A number of mixtures has been made containing coconut oil and oleo-margarine in varying proportions, and the modified Shrewsbury and Knapp process has been carried out on t'hese.The Tesults are given in Table 11. The oleo-margarine was made by mixing equal weights of premier jus and lard. Since making this mixture it has been found that the sample of lardused gives result,s which are rather above the average, but, as is considered below, this is easily allowed for.FOR THE ESTIMATION OF COCONUT OIL TABLE II.-MIXTURES OF OLEO~MAROARINE AND COCONUT OIL.Sp. Gr. 0.905 of Alco-\ 0-910 hol at 0.915 15*5*C.\ 0.920 I 77 25.0 25.3 14-2 20-5 10.8 13-3 6-2 10.6 38.7 32.2 19.5 14.4 20 1 30 1 40 - - 42.7 - 30-7 43.8 18.6 24.1 1-1- 84.2 82-7 49-9 - - 90.5 89.1 64.6 50 _./ - 61.9 32.2 60 _- 78.8 73.5 39.9 70 I 80 -I- 90 97.4 96.4 83.5 100 105.4 104.3 102.4 102.078 ELSDON AND BAGSBAWE: SHREWSBURY AND KNAPP PROCESS 94.3 93.0 92.3 77.7 When these results are plotted in the form of a curve, as is shown in the figure on p.76, it will be seen that the figures for those mixtures containing the lower amounts of coconut oil lie practically on a straight line. From this curve, knowing the Shrews- bury and Knapp value of a fat, it is possible to read off the percentage of coconut oil present. The curve is, of course, only strictly applicable to a particular sample of oleo-margarine and a particular sample of coconut oil, and the figures thus obtained will be more or less inaccurate according as the figures for the original oils are near or far away from the average; the process is, however, no worse off in this direction than any other process.As far as coconut oil is concerned, various samples appear to give uniform results; thus, six samples examined by the authors have given results varying from 101.6 to 103.6.In connection with the oleo-margarine, as has been already mentioned, the sample used by the authors is somewhat higher than that found for commercial samples of non-coconut margarine. Eight samples of such margarine have given results varying from 3.8 t o 4.7, and therefore 4.0, as the average found, seems to be a suitable figure to take.The curve, then, will have to be corrected for this difference in the base used. From the figures given for mix- tures of coconut oil, oleo-margarine, and butter, it will be seen that the shape of the curve is not altered, at any rate for the smaller percentage of coconut oil, by small differences in the composition of the non-coconut basis; we can thus get a curve, parallel to the curve obtained by plotting the figures given in Table II., from which we can read the percentage of coconut oil in commercial margarine.(b) Mixtures of Butter and Coconut OiZ.-In the same manner, a number of mixtures have been made of butter and coconut oil, and the Shrewsbury and Knapp process carried out on these.The coconut oil used was the same sample that was used for the mixtures with oleo-margarine; the butter used was a mixture of a large number of commercial samples all of which had given normal figures. The figures obtained are given in Table III., and they have also been arranged in the form of a curve (seep. 77). TABLE III.--MIXTURES OF BUTTER AND COCONUT Om.99.5 104.3 98.8 102.4 98.5 102.0 91.0 102.0 Sp. Gr. 0-910 29.8 42-4 57.3 67.0 15.0 21.0 26.8 33.0 1 18.4 24.6 32.3 44.8 , 12.0 15-6 19-8 24.7 I I , 40 74.0 56.5 44.5 31.7 50 79.0 70.1 55.3 41.0 60 ~- 84.2 79.5 68.0 52-0 70 -- 88.5 87.0 76.5 6543 80 j 90 I 100 The same remarks apply to these figures as to the figures given for the mixtures of oleo-margarine and coconut oil.Figures are given below showing the range, so far as found, for samples of pure butter. (c) Mixtures of Butter, Oleo-Margarine, and Coconut Oil.-As has been mentioned above, Revis and Bolton, in their paper, give the results that they have obtained on mixtures of butter, margarine, and coconut oil by their modification; the authorsFOR THE ESTIMATION OF COCONUT OIL 79 have to confess, however, .that they are unable to understand Table 111.as given by them. In the first place they give, as the figure for a mixture containing 75 per cent. of coconut oil, 190.6, whilst the highest possible figure that can be given by a pwre coconut oil, assuming all the acids to be soluble in the alcohol used, is about 107; it is possible, however, tha$ their figures are obtained by doubling the actual figures found by titration.In the second place the figures given by mixtures of oleo- margarine and coconut oil fluctuate in a serious manner wit,h small additions of butter fat; thus, the figure for a mixture of 25 per cent. of coconut oil and 75 per cent. of oleo-margarine is iincreased from 39.6 to 64.0 by the addition of 2 per cent.of butter fat; a further 3 per cent. of butter fat reduces the figure to 44.6, whilst a further addition of 5 per ce:nt. of butter fat increases the figure once again to 60.8. Similar results are given with the mixtures containing 50 per cent. of coconut oil. Prom these figures one can only deduce one of two things : either that the process as described does not give suficiently concordant results, or that the process is useless for such mixtures-the mixt'ures which are, after all, those for which it should be most useful. I n order to test this matter, various mixtures have been examined by the proposed new process, with, as will be seen, very different results.These results are given in Table IV. TABLE IV.-MIXTURES OF BUTTER, OLEO-MARGARINE, AND COCONUT OIL, ,ALCOHOL SP.GR. 0.920 AT 15.5" c. - - } Percentage of Coconut Oil ~ - - - _ 0.0 5-0 10.0 20.0 40.0 6000 70.0 Percentage of Butter Fat. 0.0 6.2 7.6 10.6 14.4 24.1 39.9 49.9 2.5 6.2 7 *6 10-6 14.0 24 *2 40.4 50.6 5-0 6.3 7.8 10.7 14.3 24.3 4043 52 *4 10.0 6-6 8.0 10-8 14.8 24.7 41 -8 55.3 20.0 6.9 8.2 11.0 15.0 25.8 42.5 59.6 From these figures it will be seen that the results are practically unaltered by small additions of butter fat, which is, of course, what would be expected.In the case of mixtures, therefore, in calculating the percentage of coconut oil, the presence of any amount of butter fat; up to 10 per cent. (the limit for margarines) may be almost disregarded. In the case of mixtures containing large quantities of butter- fat a small correction could easily be made.As has been stated above, the process does not give very concordant results with mixtures containing much more than 60 per cent. of coconut oil if alcohol of sp. gr. 0.920 is used; for this reason the results for the mixtures containing more coconut oil than this are given for alcohol sp. gr. 0.910 in Table V.80 ELSDON AND BAGSHAWE: SHREWSBURY AND KNAPP PROCESS ~ 2-5 78-8 84.2 90.5 97.3 TABLE V.-MIXTURES OF BUTTER, OLEO-M~RGARINE, AND COCONUT OIL, ALCOHOL SP.GR. 0.910 AT 15.5" c. 5.0 10.0 1 20.0 - 79-0 79 *2 80.3 84-2 84.8 85.8 90-6 91-8 94-3 97.9 99.5 - -~ ~ Percentage of Coconutoil } 60 70 80 90 0.0 7843 84.2 90-5 97 *4 Percentage of Butter Fat. From these tables, knowing the Shrewsbury and Knapp figures and the approxi- mate percentage of butter from the Eeichert-Polenske-Kirschner process, the amount of coconut oil present is easily calculated.The Figares f OT Pure Butter.-Fourteen samples of butter have been examined by the proposed process, and figures have been obtained varying from 9.0 to 13.0, the average being 11.0. For the present this latter figure would seem the best one to be taken in calculating the percentage of coconut oil in mixtures of butter and coconut oil, coconut oil being considered to be present when the figure is over 14.0.It is hoped that the calculation of the percentage of butter and coconut oil present together in mixtures, and a proposed method for differentiating between coconut oil and palm kernel oil, may be aealt with more fully in a pper shortly to be published.Preliminary experiments have been made with 8 view to combining the Shrews- bury and Knapp process with the Reichert-Polenske-Kirschner process. These experiments have been very successful, and it is hoped that the matter may be fully dealt. with on a future occasion. MUNICIPAL LABORATORY, S ALFORD. DISOUSSION. Mr. A. W. KNAPP said that it had been somewhat of a disappointment to Mr.Shrewsbury and himself that experienced workers had thought it necessary to modify the process, indicating that they had failed to get concordant results with the original process, because he and Mi.. Shrewsbury had succeeded very well. It was evident that a t the outset the original process presented certain difficulties, and he thought that with Messrs.Elsdoh and Bagshawe's modification workers might hope to be not so long in learning to get good results ; but at the same time it must be pointed out that the modified process was much slower than the original process, in which speed was particularly aimed at. In the course of their investigations they practically passed through the modification proposed by Messrs. Elsdon and Bagshawe, having adopted it to begin with as the most natural method; and it might therefore be interesting to look at the reasons which led them to abandon it.In the first place, the questionFOR T'HE ESTIMATION OF COCONUT OIL 81 Butter. 23.0 16-0 11.5 of the strength of the alcohol was worked out very carefully, the sp. gr. of the alcohol used being 0922, as against 0920 in the proposed modification.An earlier investigator, Robin, preferred 421 for dissolving the fatty acids of coconut oil. Vandam used a rather stronger spirit. The strength which they recommended was arrived at as follows : Coconut Oil. 216.5 190.0 173.5 TABLE I.-~OLUBILITY OF THE-FATTY ACIDS AT 25" C. Butter. -~ I. 9 22 23 To 50 C.C. Alcohol (-822) Added Water. Coconut Oil.169 208 212 36 C.C. 41 C.C. 51 C.C. Sp. Gr. of Alcohol and Water Mixture. ~922 -929 -939 Volume. 83.7 C.C. 88.5 C.C. 98.2 C.C. TABLE 11.-SOLUBILITY IN ALCOHOL (SP. GR. *922)-EFFECT OF TEMPERATURE. - Temperature. 9.5" c. 14" C. 25" C. C.C. & NaOH for Soluble Fatty Acids from 5 Grms. of Fat. lxargarine. Butter with 20 per I Cent. Coconut oil. 1 28 12 11 I 31 14 I 31 Starting with the ordinary Reichert-Meissl process, and working on coconut oil, the fatty acids obtained from the 5 grms.taken were dissolved in 50 C.C. of alcohol (industrial methybted spirit, sp. gr. *822), water being then added drop by drop until the liquid became just turbid, the idea being to find a strength of alcohol capable of holding in solution the *whole of the coconut oil fatty acids that might be present.They had thought a t one time that possibly a simple method of detecting coconut oil might be arrived at i:n that way-namely, by preparing the fatty acids from the sample, dissolving them in 50 C.C. of alcohol, and noting how much water was required just to produce hbidity. The amount of water required for coconut oil was very constant, being always about 36 to 38 c.c., but for butter it was unfor- tunately very irregular, varying from 7 to 24 C.C.That, however, was how the 36 C.C. of water used in the process was arrived at, giving an alcohol wifha sp. gr. of -922, and this was the strength which they had found to give the widest range between82 ELSDON AND BAGSHAWE: SHREWSBURY AND KNAPP PROCESS coconut oil and butter.In their earlier experiments they worked similarly to the pment authors, but only allowed ten minutes' standing; but on making experiments with pure glycerides, they found that there was, at ordinary temperatures, a tendency fo form supersaturated solutions. The question was one of difEerentia1 solubility, and they found that, at temperatures below 14" C., the solubility of the fatty acids of coconut oil rapidly decreased, whereas above 14" it increased only very slightly.A higher temperature was therefore tried, and as a result the fatty acids were ob- tained in a liquid state, so that, while those undissolved floated to the top, the liquid containing the dissolved coconut oil fatty acids could be run off from below. They had always recognised that the whole of the fatty acids did not rise in the three minutes, but it was found that when a longer time was allowed (say thirty minutes) the dserence between the figures for coconut oil and for butter was decrmsed.In Messrs. Elsdon and Bagshawe's process he thought it important that the tempera- ture should not be allowed to fall below 15.5" C. For large quantities of coconut oil the process did not even yet seem to be satisfactory, but he gathered that deci- normal solutions were used for titration, which would involve the addition of a large quantity of water to the alcoholic solution, and thus spoil the end reaction.Better results would probably be obtained if in titrating one either added more alcohol or, which would be preferable, used normal alkali for titration.Mr. C. REVIS said that if Messrs. Elsdon and Bagshawe had done nothing more, they would have done good service to anyone who had to deal with fractional crystal- lisation of fatty acids from alcoholic solutions, in showing how greatly the results were affected by the strength of the alcohol, and how difficult it WEM to escape from the effects of supersaturation. This had also been shown in some recently published American work in connection with Hehner and Mitchell's method of estimating stearic acid. Although the authors had obtained excellent results from the scientific point of view, he feared that from the practical laboratory point of view the process stood condemned on the very tables given in the paper.From the large differences caueed by comparatively small variations in the strength of the alcohol, it was evident that the necessary delicacy of manipulation could not be maintained under ordinary conditions.Mi. Bolton and he did not intend to throw any aspersion on the original process; it was not uncommon for a process satisfactory in the hands of its author to be wrongly interpreted by other workers. When Mr.Bolton and he starfed to investigate this subject, they were under the urgent necessity of finding a quick and accurata way of arriving at the composition of these coconut oil mixtures, which then were just coming into use, and they thought it well to begin by examin- ing existing processes rather than to devise a new one. They found that the standad procesa which everyone used-the Reichert-Meissl-Polenske process-was capable of doing what was needed as accurately and as well as any other method when carried out in a standard way, and, since the figures obtained by Messrs.Shrewsbury and Knapp's process appeared to them to give less reliable information than that given by the standard process, it did not seem worth while to investigate it further. Mr. KNAPP said that at the time when he and Mr.Shrewsbury devised their process the Polenske method was not fully appreciated, and he agreed that, pro-FOR !PEE ESTIMATION QF COCONUT OIL 83 Reichert- Meissll. 29.5 3.6 vided the coconut oil was not specially prepared, the Polenske process was sufficient for the purpose. It was possible, however, so to prepare coconut oil products that the amount present would not be disclosed by the Polenske process, and it was then that a process depending; on alcohol solubility became useful, since it depended, not on the volatile fatty a,cids, but on the main bulk (60 per cent.or so) of the COCO- nut oil itself. As an example, the figures on a, butter containing 20 per cent. of a commercial coconut " stearine " are of interest.Polenske. 2.9 7.2 Butter A . . .. .. .. Coconut stearine . . . . .. Butter fat A with 20 per cent. coca- nut stearine . . .. .. I 24-4 1 4.0 Kirschner. 21.5 0.3 17.2 Shrewsbury and Knapp. 32-6 163 -5 40-3 It will be seen that the Polenske process indicates the presence of only 10 per cent. of coconut oil, whilst the Kirschner indicates the absence of about 20 per cent. of butter, and the Shrewsbury-Knapp process the presence of 17 per cent.of a coconut oil product. In order to test the relative difficulty of obtaining duplicates when first trying the process, Mi. R. V. Wadsworth, who had never performed either process, tried both the Polenske and our process. On the same butter he obtained the following results in the order given.. Polenske figure: 2.6, 2-9, 2.9, 3.0; Shrewsbury-Knapp figure: 346, 32.3, 31.5, 32.3.It may be pointed out that for calculation purposes an error of 0.1 C.C. in the llolenske is equivalent to 0.8 to 1.0 C.C. in our process. Mr. BOLTON thought that, provided the coconut oil was satisfactory as regards taste and absence of free fatty acid, the manufacturer would be content, and would not trouble much whether its presence could be detected or not.Moreover, manu- facturers of coconut oil usually dealt also with palm-kernel oil, and the two often got mixed. A process for the detection of palm-kernel oil had been devised by Burnett and Revis (ANALYST, 1913, 38,255). When palm-kernel oil, and also other lesser known oils of the palm-kernel family, happened to occur in margarine, abnormal figures would be obtained, but that did not mean that the coconut oil had been specially prepared, having a composition calculated to defeat analytical tests.84 QUANTITATIVE ESTIMATION OF MERCURY IN ORGANIC COMPOUNDS THE QUANTITATIVE ESTIMATION OF MERCURY IN ORGANIC COMPOUNDS.BY 3. E. MARSH, M.A., F.R.S., AND 0. G. LYE. (&ad at the Meeting, blQrch 7, 1917.) THE process here described is a modification of the method of estimating mercury by combustion with quick-lime. When organic mercury compounds, especially aromatic derivatives, are analysed by the ordinary lime-combustion method, there occurs frequently, besides a deposit of carbon on the lime, a distillate of tamy or crystalline substances. These mix with the mercury and make it difficult to purify. We have found that the mercury derivatives of aromatic sulphonic acids, amlysed in this way, gave no condensable distillate other than mercury. The lime, for the most part, remains uncoloured, and calcium sulpfiide is formed by the reduction of the sulphonic group. We have also found that for mercury organic compounds other than sulphonic acids it is only necessary to mix the substance with a sulphate such as calcium sulphate and with quick-lime to obtain a very good result, the mercury being collected in a single globule without any tarry or crystal- line distillate. The process then consists in distilling the mercury compound mixed with about twice its weight of dry calcium sulphate and with excess of quick-lime in the ordinary way. To expel the mercury vapour left in the tube at the end of the combustion we have found it best to use carbon monoxide evolved from calcium oxalate which is placed a t the closed end of the combustion tube. With one com- bustion furnace two analyses can be carried out at the same time, one a t each end of the furnace, and finished in less than an hour,