288 Analyst, April, 1971, Vol. 96, pp. 288-295 The Determination of Glycerol by the I.U.P.A.C. Form of the Malaprade Method* BY R. F. BARBOUR (Newcastle Technical Centre, Procter 6. Gamble Ltd., Newcastle upon Tyne) AND J. DEVIm ( Unilever Research Laboratory, Unilever Ltd., Port Sunlight, Cheshire) The accepted periodate method for the determination of glycerol has come under review by the International Organisation for Standardisation (Sub-committee ISO/TC47/GT2) and certain modifications have been sug- gested with regard to the different pH end-points for the sample and blank and the possible loss of formic acid from the system by volatilisation. The present paper summarises the view of the U.K. delegates to ISO/TC47/GT2, which are: (i) The choice of pH about 8.0 for the sample is justified on the ground that it contains formic acid in addition to strong acids.The blank, which contains only strong acids, should, on general grounds and by calculation, be titrated to pH 7-0. The use of pH 6.5 instead of 7.0 for the blank is an empirical correction designed to compensate for some lack of stoicheiometry or other bias in the procedure and to bring the results into agreement with those obtained by independent moisture and specific gravity determinations. This “correction” amounts to 0-03 per cent. (ii) There is a potential loss of formic acid by volatilisation from the system (estimated variously to be equivalent to 0.01 to 0.04 per cent. of glycerol), which partly explains the apparent lack of stoicheiometry. (iii) Modifications have been proposed to minimise this loss of formic acid and (by adding formate ions to both solutions) to unify the end-points at or about pH 8.0.Unless these modifications lead to a pronounced improve- ment in reliability, it seems doubtful whether further extensive trials on an international basis would be justified. THE accepted method of determining glycerol is based on oxidation by sodium periodate to formic acid and measurement of the latter by titration with standard alkali, which is a modification of that outlined by Malaprade.1 It involves the use of an end-point of pH 6.5 for the blank and of pH 8.1 for the sample. This procedure was originally put foward by the American Oil Chemists’ Society and was substantiated by the results of a series of inter- national collaborative trials conducted under its auspices in 1956-57.It was subsequently adopted by the International Union of Pure and Applied Chemistry (I.U.P.A.C.)2 and also by the British Standards Institution (B.S.I.) and other national bodies, and as such it has formed a satisfactory basis for commercial transactions in glycerine in all parts of the world. More recently, the method has come under review by the International Organisation for Standardisation (Sub-committee ISO/TC47/GT2), when it was stressed that the choice of pH values for the dual end-points has not been adequately clarified. The belief was expressed that this discrepancy existed to compensate for some consistent error or bias elsewherein the test, which should be further examined before adoption.As a result of experimental work by members of the Sub-committee, the following suggestions have been made: there * This paper was written a t the suggestion of the British Standards Institution Committee CIC/6- Glycerol by two of its members who are also U.K. delegates to the Working Group on glycerol of the International Organisation for Standardisation (ISO/TC47/GT2). It therefore represents the current U.K. attitude on the problem of the determination of glycerol, which is a t present engaging the attention of ISO. 0 SAC and the authors.BARBOUR AND DEVINE 289 is a slight loss of formic acid from the test sample by volatilisation, which could be a source of consistent erro1-39~; and a unified end-point could be achieved by addition of formate ions to both sample and blank.4 It is appropriate, therefore, to record the views of the U.K.delegates to the committee, which are summarised as follows. pH OF END-POINTS CURRENTLY USED- In the standard I.U.P.A.C. procedure, excess of periodic acid (a weak acid) is reduced before titration to iodic acid (a strong acid) so that its buffering action is eliminated. The acids being titrated are, therefore, for sample: sulphuric, iodic and formic acids, and for blank : sulphuric and iodic acids. Nominally, therefore, the end-points used for the titrations (pH 8.1 and 6.5) should accord with the equivalence points of formic acid and of the sulphuric acid - iodic acid mixture, respectively. It is conceded that it is unusual to subject a blank to a different procedure from that used for the sample and for this reason, although we retain the term here, it might be better referred to as a control.In the 1956 deliberations of the B.S.I. Sub-committee on Glycerine it was suggested that the result of carrying the pH of the sample to 8.1 was to over-titrate the strong acids contained therein, and that the blank should therefore also be titrated to pH 8.1 so that the errors in each instance should cancel out in the subsequent calculation. This suggestion was rejected on theoretical grounds and was not supported by practical tests in which solutions of formic acid and of a mixture of formic, sulphuric and iodic acids in concentrations equivalent to those of an actual glycerol determination were titrated potentiometrically under nitrogen with 0.1 N carbonate-free sodium hydroxide solution from pH 6.5 to 8.1.There was no significant difference in the amounts of alkali required by the two solutions unless there was contamination by atmospheric carbon dioxide (Lazarus, W., private communication), thereby indicating that in the mixture the alkali required to cover this pH range is determined effectively by the formic acid alone. No additional alkali is required by the presence of sulphuric and iodic acids, which are therefore not over-titrated. It is appropriate to consider what end-points for sample and blank might be predicted by strictly theoretical means. These are recorded in Appendix I, which gives estimates of the pH at the stoicheiometric end-points for solutions containing, for sample : sodium formate, iodate and sulphate, and for blank : sodium iodate and sulphate.The calculated pH values are 7.96 for the sample and 7.05 for the blank, respectively. This confirms the need for two different pH end-points for sample and blank and reasonably supports the choice of pH 8.1 for the sample titration, but not pH 6.5 instead of about 7 for the blank. Examination of the original correspondence in the period 1952 to 1955 between the United Kingdom Glycerine Producers' Association and the American Oil Chemists' Society reveals that the A.O.C.S. Glycerine Analysis Committee tried various pH values between 6.5 and 7-5 for the blank on a largely empirical basis; of these, pH 6-5 was selected as giving the best results, i.e., those most in conformity with the independent assay of the sample by specific gravity and moisture determinations.This latter method of assay was based on the belief that the sample of glycerol for assay was pure and conformed with that used for establishing the standard Bosart and Snoddy tables of specific gravity for glycerol - water solutions.5 The pH 7.0, which was the inflexion point actually found for the titration curve," was discarded, and as late as 1957 pH 6.5 was being erroneously quoted as the equivalence point. It is clear, therefore, that the pH of about 7.0 as the equivalence point for the blank is borne out by both theory and practice and that the selection of pH 6.5 in its place represents an empirical correction designed to allow for some slight imperfection in the over-all analysis, whether of stoicheiometry or otherwise.The effects of such pH differences as have been quoted are indicated by the p H - titration curves presented in Appendix B of the Minutes of the October 1965 Meeting of ISO/TC47/GT2, from which the following figures are taken. * It is recognised that inflexion point and equivalence point are not identical, but with strong elec- trolytes, as in the blank, the difference is negligible.6290 BARBOUR AND DEVINE : DETERMINATION OF GLYCEROL BY [Analyst, Vol. 96 Blank-A pH of 6.5 to 7-0 requires 0.012 ml of 0-125 N alkali solution equivalent to 0.03 per cent. of glycerol (at 0.41 g sample weight), and a pH of 6.5 to 8.1 requires 0.036 ml of alkali solution equivalent to 0-10 per cent. of glycerol. Sample-A pH of 6.5 to 7.0 requires 0.08 ml of alkali solution equivalent to 0.22 per cent. of glycerol, and a pH of 6.5 to 8.1 requires 0.16ml of alkali solution equivalent to 0.45 per cent.of glycerol. Thus, titration of the blank to pH 6.5 instead of 7.0 has the effect of raising the glycerol result by about 0.03 per cent. However, titration of both sample and blank to pH 7.0 introduces an error of about -0.26 per cent. of glycerol,* while titration of both to pH 8.1 introduces an error of about -0.10 per cent. of glycerol. In this consideration it is important that no significant amounts of other buffers should be present, e.g., carbon dioxide from the atmosphere or impurities in the sample. USE OF MODIFIED END-POINTS- As the presence of formic acid in the sample titration has the effect of raising the stoicheiometric or “equivalence” end-point to about pH 8, it was suggested that formic acid (or sodium formate) should be added to both blank and sample, which should be then titrated to pH 7.9 O*Z4 This addition would eliminate the need for different end-points and yield titration curves of similar character.Tests carried out in three laboratories, in which the results obtained in this way are compared with those by the normal I.U.P.A.C. method, are recorded in Appendix 11. They confirm that, within experimental error, the modification yields the same results as the normal blank to pH 6.5. This seems to remove the apparent undesirability of using different end-points for sample and “blank.” However, errors can arise in the procedure of adding the formate required to unify the end-points and more extensive tests would be necessary to assess the effect on reproducibility.VOLATILITY OF FORMIC ACID- Mormont and Gillet3 have pointed out that because of the exothermic reaction between glycerol and periodate there is a slight rise in the temperature of the solution and consequently some evaporation: this effect is made obvious by the slight condensation that occurs on the under surface of the clock-glass with which the beaker is closed. The condensate from the sample is acidic in contrast to that from the blank. This acidity led to the suggestion that there is a loss of formic acid from the vessel by volatilisation and that this loss is the source of the systematic error that is otherwise accounted for by the use of the empirically chosen value of pH 6.5 for the blank titration.In a model solution of formic and sulphuric acids similar in composition to that in a glycerol determination, a loss of 0.2 per cent. of its strength was demonstrated a t ambient temperature within 30 minutes in the open air (‘‘2 l’air libre”). This problem was examined in the U.K. in 1953 (Lazarus, W., Walley, G., and Wilkie, A. L., private communication) when an equivalent solution of formic acid was titrated, with and without a stream of nitrogen being bubbled through the solution for periods of up to 20 minutes prior to and during the titration: no significant loss of formic acid was detected. In a series of tests, variations observed in carrying out the glycerol determination at 0, 20 and 45 “C with the procedure then in use were not regarded as significant, and variations for a slight rise above room temperature certainly could not be regarded as significant. We have re-examined this problem in the light of Mormont and Gillet’s observation, taking into account the finer differences now being looked for, with the results given in Appendix 111.We agree that there is a rise of 4 to 5 “C in the temperature of the solution and have shown that the presence of formic acid can be specifically demonstrated by a colorimetric test in the condensate on the clock-glass from a single determination.’ While there is no doubt about the potential volatilisation of formic acid, the extent of the loss from the system in the average analytical procedure is more difficult to assess. Our recent tests indicate losses varying from zero up to 0.04 per cent.of glycerol, although some uncertainty must attach to those tests in which the solutions were left open to the atmosphere to exag- gerate the effect; the subsequent loss would be compensated for to some extent by absorption * This is close to the figure of “about 0.370” given in B.S. 2621-5 : 1964 as the correction that results from using the pH end-points 6.5 and 8.1. In our view the term “correction” should more properly be applied to the 0-03 per cent. difference that results from using pH 6.5 instead of 7-0 in the blank.April, 19711 THE I.U.P.A.C. FORM OF THE MALAPRADE METHOD 291 of atmospheric carbon dioxide. This tendency towards compensation must, of course, operate whatever the imperfection in the closure of the beaker may be.The potential volatilisation of formic acid has been re-examined also by Mormont, Gillet and Hei~~erth.~ The condensates from several tests were collected and tested by infrared spectroscopy, which again demonstrated the presence of formic acid. The con- densates on the clock-glasses from thirty-eight tests were also collected and combined for titration; the result obtained was equivalent to a mean recovery per test of 0.01 per cent. of glycerol on the clock-glass. The actual Loss by leakage from the system was thought to be in excess of this value, by a factor approaching 10, and the difference between the loss of formic acid and the gain in carbon dioxide, for a model mixture of formic and sulphuric acids, has been put at the mean figure of 0.2 per cent.per hour (Mormont, R., private com- munication). However, some doubt exists on whether this refers to open air conditions (“& l’air libre”) or conditions with a closed beaker (“en becher couvert”) since both these forms are used. The loss of formic acid from the system, found in terms of glycerol by two sets of investi- gators therefore varies from zero to upwards of 0.04 per cent. with one set, and approaches 0-2 per cent. with the other. The mean effect in our own tests is of the same order as the 0.03 per cent. of glycerol, which is equivalent to the gain accounted for by titrating the blank to pH 6.5 instead of 7.0. However, the following considerations make it inadvisable to equate the two effects at present: (a) it cannot be said whether it is normal practice to wash down the clock-glass before titration; (b) the formic acid in the condensate is not a measure of that lost from the system by volatilisation; and (c) the effective loss by volatilisation depends on: (i) the type of beaker (600m1, tall, in I.U.P.A.C.directions; 600m1, squat, in B.S. 2621-5: 1964), (ii) fit of the clock-glass, (iii) size of the beaker lip and (iv) absolute reaction temperature reached, which will depend partly on the initial ambient temperatures of the laboratory and of the reagents. It may be thought that there is reason to reconsider the equipment used, because originally the use of a beaker was presumably conditioned by the necessity of inserting separate elec- trodes.Present-day electrodes can be inserted as a single piece through a small ground-glass joint so that a suitable standard flask could be used with effective stoppering duringthe reaction period. Alternatively, a beaker without a lip and the top surface ground to take a ground-glass plate could be used. Even with the desire to retain as simple equipment as possible, the failure to close the vessel, in the light of errors ascribed to the passage of volatile components, appears to be a retrograde step. This still leaves a minor problem as to how much formic acid remains, after cooling for half an hour, in the vapour phase, some of which might disperse on opening the beaker to the atmosphere to conduct the titration. This amount should not be significant, especially if the simple expedient of immersing the reaction vessel in a bath of cold water is adopted.* The alternative approach chosen by Mormont et aZ.* is to damp down the exothermic effect, not by external cooling, but by adding the dilution water before the periodate reagent instead of at the end of the reaction.This is claimed to halve the error caused by the combined loss of formic acid and gain of carbon dioxide. However, it appears to be generally held by analysts familiar with the collaborative work in the U.K. in the 1950s that over-dilution affects the determination unfavourably, possibly because of the difficulty of excluding carbon dioxide, which affects the sample and blank to different extents. CONCLUSIONS The following aspects have been clarified by recent work on the standard procedure for the periodate determination of glycerol : (i) The use of two different pH end-points for the blank and sample is necessitated by the different nature of the acids being titrated, but whereas a pH of 8.1 for the sample is in reasonable conformity with the equivalence point of the formic acid, which is present only in the sample, the pH required for the blank, both as determined and for theoretical reasons, is about 7.0.The use of a pH of 6.5 in its place is not warranted except as a correction to account for an apparent lack of complete stoicheiometry in the reaction. This form of * Our calculation suggests a total vapour content equivalent to less than 0.02 per cent. of glycerol, under present standard conditions.292 BARBOUR AND DEVINE : DETERMINATION OF GLYCEROL BY [Autalyst, Vol.96 correction does not commend itself to all analysts; others accept it on the grounds that it is equivalent to only 0.03 per cent. of the glycerol present and that a correction of this magni- tude hardly warrants modification of an established method that has served industry well. (ii) There is a potential loss of formic acid from the system by volatilisation which, in principle, could account for the apparent lack of stoicheiometry. On the basis of the restricted amount of evidence available it is not yet possible to equate these two factors. (iii) Modifications to the standard procedure have been suggested4 to overcome these difficulties. They entail (a) adding the dilution water before instead of after the reaction, and (b) adding equal amounts of sodium formate to blank and sample, both of which are then titrated to the equivalence point of the formic ions, which is considered to be pH 7.9 0.2.The few tests carried out on item (b) give the same results as the standard I.U.P.A.C. pro- cedure, which it is designed to supplant as being based on sounder theoretical principles. Further extensive trials are envisaged to determine which of these procedures is preferable on grounds of accuracy and reproducibility. (iv) It is questionable whether further work to detect a bias of 0.03 per cent. of glycerol in a test with a standard deviation between laboratories that we estimate to be of the order of 0-22 per cent. is justifiable, as a very large number of tests would be required.But if this was thought desirable we suggest that an alternative approach, which should be considered, would be prevention of any formic acid loss by slight modification of the apparatus. This would only entail effectively sealing it and placing it in a bath of cold tap water during the reaction. It would then be appropriate to use the true equivalence points of pH 7.0 and about 8-0 for the respective titrations. This would have the advantage of placing the method on a sound theoretical basis without altering the nature of the solutions to be titrated and possibly introducing errors in the addition of sodium formate. The authors thank Procter and Gamble Limited and Unilever Limited for permission to publish this paper and for the necessary facilities provided. REFERENCES 1.2. 3. 4. 5. 6. 7. Malaprade, L., Bull. SOC. Chim. Fr., 1928, 43, 683. Standard Methods of the Oils and Fats Section of the I.U.P.A.C., Fifth Edition. Buttenvorths, Mormont, R., and Gillet, A. C., jun., XXXVIe Congrks Int. Chim. Ind., Bruxelles, September Mormont, R., Gillet, A. C., jun., and Heinerth, E., Talanta, 1969, 16, 701. Bosart, L. W., and Snoddy, A. O., Ind. Engng Chem., 1927, 19, 506. Marinenko, G., and Champion, C. E., Analyt. Chem., 1969, 41, 1208. Feigl, F., “Spot Tests in Organic Analysis,” Fifth Edition, Elsevier Publishing Company, Amster- Received August 13th, 1970 Accepted November 12th, 1970 London, 1966. 1966, Gr. X. S26, 691. dam, London, New York and Princeton, 1956, p. 340. Appendix I CALCULATION OF END-POINTS FOR SAMPLE AND BLANK Solution volume = 300 ml Periodate added = 3.0 g Molar concentration of iodate formed therefrom Glycerol taken = 0.41 g (at 100 per cent.glycerol) Molar concentration of formic acid (and formate) produced Dissociation constants at 25 “C, taken from the “Handbook of Chemistry and Physics,” Forty-fifth Edition, 1964 (Chemical Rubber Co.), are formic acid 1.77 x 10-4 = K,; iodic acid 1.69 x 10-1 = Kb; and water 10-14 = K,.April, 19711 THE I.U.P.A.C. FORM OF THE MALAPRADE METHOD 293 HB + B- + H+ + A- + HA With sodium formate and iodate present the equilibrium may be represented as- I t OH- where A- and B- are formate and iodate ions, respectively. Where Ca and c b are total molar concentrations of (formate ions + formic acid) and (iodate ions + iodic acid), respec- tively, this equilibrium gives- [H+] x [OH-] = Kw [H+] x [A-] =Ka [HA] [H+] X [B-] = Kb [HB] [B-] + [HB] = Cb = 0.047 [A-] + [HA] = Ca = 0.015 [H+l + [HA1 + [HBI + [H,OI = [OH-I + [H,OI.(In these dilute solutions the activity coefficients are taken as close to unity.) This last relationship is based on the fact that when only pure sodium formate and iodate are present, without added acid or alkali, the “free” plus “combined” H+ ionic concentration must equal the “free” plus “combined” OH- ionic concentration, which leads to the following relation- ship- [H+I4 + [H+I3 (Ka + Kb + ca + cb) + [H+I2 (KaKb + KaCb + KbCa - Kw) - [H’] (KaKw + KbKw) - KaKbKw = O . . .. .. * . (1) or, applying the above numerical values (2) A Sturm analysis indicates only one positive root ; and this may be approximated, iteratively, by initially ignoring the 3rd and 4th degree terms, solving the remaining 2nd degree relation for the positive root, inserting this in the 3rd and 4th degree terms, absorbing their values in the constant term, re-solving, re-inserting and proceeding to convergence.[H+I4 + 0-23 [H+I3 + 2.6 x [H+I2 - 1.7 x [H+] - 3.0 x 1O-l’ = 0 . . Thus, the 2nd degree solution is 1 1 Kw2 (Ka + Kb)2 [H+l = 2 (KaKb + KbC, [ KW (Ka + Kb) + + 4 KaKbKw (KaKb -t- KbCa + KaCb - Kw) -k KaCb - Kw) In this instance Kb is of the order of 1 000 x Ka; and hence in the above bracketed sums the only significant terms are those in Kb and KbCa. The above thus reduces to 1 = 1.1 x IH + 1 [ ~ w + J ~ w 2 + 4 ~wKa~a.1 - 2 c a applying the above numerical values. This gives pH = - log [la1 x 10-5 = 7.96.Inserting the above value in the 3rd and 4th degree terms yields a value of the order while the constant term is of the order 1O-l’. Thus the 3rd and 4th degree terms have negligible effect on the estimate, which remains at pH 7-96; and this indicates that a suitable pH end-point for the sample titration would be expected to be about 8. In the absence of iodate ions, c b = 0, and [H+] + Kb becomes a factor in equation (l), thereby cancelling and reducing this equation to- As before, ignoring the 3rd degree term, this yields- [H+I3 + [H+I2 (Ka + Ca) - [H+] Kw - KaKw = 0 = 1.1 x 10-8, applying the above numerical values.294 BARBOUR AND DEVINE: DETERMINATION OF GLYCEROL BY [Analyst, Vol.96 As before, the 3rd degree term remains negligible. This gives pH = 7-96, which is effectively the same as that when the iodate is present with the formate, as is to be expected from the form of the two expressions, where Ka is approximately 100 times smaller than Ca. Thus the iodate has practically no effect on the pH produced by the formate and, for the same reason, the effect of the sulphate can be ignored in the sample and blank titrations. In the blank titration formate is absent and, as above, the relationship becomes with Ca = 0, 1 r I 1 = 0.89 x 10-7, applying the above numerical values. Thus a suitable pH end-point would be expected to be about 7. This gives pH = -log (0.89 x lo-’) = 7.05. Appendix I1 COMPARISON OF I.U.P.A.C. PROCEDURE WITH THE “UNIFIED END-POINT” PROCEDURE The data are taken from document ISO/TC47/GT2-No.87; a record of tests performed in Brussels, Newcastle and London is given as follows. The method of test consisted in performing the normal I.U.P.A.C. glycerol determination to the usual end-points, viz., sample pH 8.1 and blank pH 6.5, at which stage equal amounts of formic acid solution were added to both sample and blank, and the titrations were con- tinued, both to pH 8.1. For this second stage the sample and blank titrations are taken as the sums of the titrations before and after formic acid addition, respectively. Bvussels*- Sample: to pH 8.1 Blank: to pH 6-5 to pH 8.1 (Sample less blank) (Sample less blank) (Sample less blank) Titration/ml I A \ (11) With added formic acid (1) I.U.P.A.C... . . . . 29-90 29-91 45.99 46.00 .. .. .. 4.89 4-89 - - . . . . 20.97 20.98 25.02 .. . . . . 25-01 25.02 (Mean). . . . 25.015 25.02 (Mean) (I - 11) - .. 25-02 - 0.005 For the amounts of sample taken this is equivalent to -0-018 per cent. of glycerol. * Reported by Mr. A. C. Gillet, jun., of Solvay et Cie, S.A., to whom our thanks are given. Titrationlml I L \ (11) With added formic acid London- (1) (with distilled glycerine) I.U.P.A.C. Sample: to pH 8.1 . . .. .. .. . . 42.65 62.37 topH 8.1 . . . . . . .. .. 24.56 (Sample less blank) . . . . . . .. .. 37.81 37.81 4.84 - Blank: to pH 6.5 . . . . .. .. .. (Sample less blank) (I - 11) . . . . . . 0.00 - Titration /ml r A 1 (11) With added formic acid Newcastle- (1) (three crude glycerines were used) Sample: to pH 8.1 .. . . . . 36.64 38.06 37.54 62-01 63.45 62.96 I.U.P.A.C. Blank: to pH 6-5 . . 4.78 4.78 4.76 4.78 - - - - . . . . to pH 8-1 . . . . . . - - - - 30.14 30.20 30-18 30-20 Mean blank . . . . . . . . 4.775 30-18 (Sample less mean blank) . . . . 31.865 33.285 32.765 31-83 33-27 32-78 (Sample less mean blank) (I - 11) + 0.035 +Ow015 -0.015 (Sample less mean blank) (I - 11) (Mean) +0.012 For the amounts of sample taken this is equivalent to +0-028 per cent of glycerol.April, 19711 SUMMARY- THE I.U.P.A.C. FORM OF THE MALAPRADE METHOD 295 Sample less blank (I - 11) (Mean) /ml Brussels . . . . . . - 0.005 London . . . . . . 0.00 Newcastle . . . . . . +0.012 Mean . . . . + 0.002 This indicates that, within experimental error of measurement, titration of the blank to pH 6.5 gives effectively the same result as addition of formic acid and titration of the blank to pH 8.1.This does not, however, measure the resulting reproducibility. Appendix I11 VOLATILISATION OF FORMIC ACID DURING THE DETERMINATION OF GLYCEROL (i) The condensate on the watch-glass from a single determination was treated with magnesium powder and dilute hydrochloric acid and then with sulphuric acid and chromo- tropic acid according to Feigl.' Production of a violet -pink colour indicated the presence of formic acid in the original sample." The following data are taken from Document ISO/TC 47/GT2-No. 66: (ii) The I.U.P.A.C. glycerol determination was carried out and the traces of condensate on the clock-glasses were collected separately and titrated. Two such tests showed an average acidity equivalent to 0.02 ml of 0.125 N sodium hydroxide, i.e., equivalent to about 0-04 per cent. of glycerol. (iii) The loss of forrnic acid to atmosphere from the test-reaction liquor was measured over 72 hours, and two such tests showed an average loss per hour (i.e., approximate duration of normal test) equivalent to 0.025 per cent. of glycerol. The loss during an actual deter- mination would probably be slightly higher, because during the first 0.5 to 1 hour the liquor is slightly warmer than it is later. (iv) Three tests similar to (ii) above, but with the condensate washings combined, showed a total equivalent to 0.02 ml of 0.125 N sodium hydroxide. This is equivalent to about 0.015 per cent. of glycerol per test. (v) A stream of nitrogen was passed through an aqueous solution of formic acid, of concentration similar to that in a glycerol determination, and the acid was subsequently titrated. No significant loss of formic acid was found. The above tests indicate that during the glycerol determination there could be an actual or potential loss of formic acid equivalent to 0.015 to 0-04 per cent. of glycerol. Stevenston, Ayrshire. * We are indebted for this test to Mr. S . M. Farrer of Imperial Chemical Industries Ltd., Nobel Division,