What has actually been measured inside a fermenting heap or box: which microbes, in what order, doing what — and how much of the textbook account rests on a handful of studies.
26 studies1994 to 2025Each checked against its published abstract
What these studies add up to
The sequence of yeasts, then lactic acid bacteria, then acetic acid bacteria is well supported, and the same few species recur across countries.
Which of them are necessary is disputed. Controlled laboratory fermentations made acceptable chocolate with the lactic acid bacteria suppressed, and not with the yeasts suppressed; a 2020 review of the field still assigns lactic acid bacteria several roles, including producing the lactate on which the acetic acid bacteria grow.
The chemistry that makes cocoa flavour possible happens inside the bean, driven by acid and heat arriving from the pulp, and was worked out in vitro in the 1990s.
A systematic review found no specific relationship between the microbes present and either geography or fermentation method.
Fermentations that look alike do not taste alike. Seven Ghanaian heaps with the same microbes gave different chocolates, and what was left of the bitter compounds after fermentation set how bitter each was.
Fermentation does not finish the job. Drying can undo it: slow drying in shade let mould and putrefaction in, and one oven-drying trial left the acids in the bean and gave a tainted chocolate.
Cocoa aroma precursors are made by two enzymes with different acid optima
How are the precursors of cocoa aroma formed inside the bean?
How
In vitro digestion of cocoa seed proteins by the seed's own enzymes at controlled acidity.
Finding
An aspartic endoprotease and a carboxypeptidase, with different pH optima, act in turn; the aroma precursors are among the peptides and amino acids they release.
Does not establish: How closely a real fermentation follows the conditions that gave the best result in the test tube.
Limits: Done in vitro on prepared seed powder, not in a fermenting bean.
How does the drying method affect acidity and volatile fatty acids in the beans, and the flavour of the chocolate?
How
Beans from two clones, pod-stored for nine days and fermented for five with one turning, were dried four ways: in an oven at 60 °C, by blown air, in shade and in sun. Acidity and volatile fatty acids were measured and dark chocolate was made and tasted.
Finding
Sun-dried and air-blown beans did not differ in acidity or in the heavier volatile fatty acids, and both were judged favourably. Oven-dried beans held a high concentration of volatile fatty acids and gave chocolate with a strong off-flavour. Wet beans kept in shade for 120 hours did not dry fast enough to stop mould and putrefaction, and oven-drying them afterwards did not rescue the flavour.
Does not establish: That artificial drying always harms flavour. Other entries here found drying at 70–80 °C comparable to sun drying on volatile profile.
Limits: Two clones mixed, one fermentation protocol, one oven temperature.
Which bacteria carry out spontaneous heap fermentation in Ghana, and what do they produce?
How
Heap fermentations in Ghana followed with both culture-based and DNA-based identification and with measurement of metabolites.
Finding
Limited biodiversity. Four main clusters of lactic acid bacteria and four of acetic acid bacteria were found, with Lactobacillus plantarum, Lactobacillus fermentum and Acetobacter pasteurianus apparently playing the significant roles.
Does not establish: That those species are required, only that they were the ones present and active.
Limits: A small number of heaps in one country, and heaps are not uniform inside.
What do turning and the surrounding environment do to a heap fermentation?
How
Six spontaneous heap fermentations in Ghana, turned and unturned, with microbiology, metabolite analysis and sensory analysis of the resulting chocolate.
Finding
Turning increased aeration, the relative population of acetic acid bacteria and the production of acetic acid, and gave a more sour chocolate. The microbes were the same on farm and at a factory site, pointing to the pod surface as the main source of inoculum.
Does not establish: That turning is bad. It shows turning has a cost in acidity at this scale, not what it does in a large box.
Limits: Six small heaps. The result is about heaps of that size.
Seven heaps, the same microbes, and chocolates that tasted different
Camu and colleagues, 2008. Journal of the Science of Food and Agriculture. doi:10.1002/jsfa.3349
Question
Do differences between one farm's fermentation and another's show up in the chocolate?
How
Seven independent spontaneous heap fermentations in Ghana, on different farms and in different seasons, were followed for microbes and metabolites, and chocolate was made from each.
Finding
The same micro-organisms were involved in every heap, yet temperatures and metabolite concentrations differed, and the differences in microbial activity could be linked to the flavour of the chocolates. Epicatechin and theobromine fell during fermentation, and what remained set how bitter the chocolate was.
Does not establish: Which difference in a heap produces which flavour. The abstract's own conclusion is that control of fermentation may direct flavour, not that it has been shown how.
Limits: Seven heaps in one country. The link between microbial activity and flavour is an association across those heaps.
At which stage of farm processing do ochratoxin-producing moulds and the toxin itself appear in cocoa?
How
222 samples of cocoa from Bahia, Brazil, taken at stages of fermentation, drying and storage, tested for moulds and for the toxin.
Finding
The toxin was found at all stages, with the major incidence during drying and storage. Levels were in general low, and correlated strongly with one mould species.
Does not establish: Levels elsewhere, or a risk to anyone eating chocolate.
Limits: One producing region, Bahia, in the seasons sampled.
Can a defined starter culture control cocoa fermentation on working farms, and what does the chocolate taste like?
How
Two starter mixtures, one of a lactic and an acetic acid bacterium and one adding a yeast, were applied to heaps and boxes on farms in West Africa and Southeast Asia, twelve fermentations in all, and chocolates were made.
Finding
The inoculated species outgrew the wild microbes and prevailed in every fermentation. Both mixtures gave chocolates the authors describe as reliable in flavour whatever the region or method. For a standard bulk chocolate the yeast was needed, to give consistent ethanol production.
Does not establish: That a starter culture makes better chocolate than a well-run spontaneous fermentation, or that it pays at farm scale.
Limits: Twelve fermentations. 'Reliable' is the authors' word, and the abstract gives no sensory scores or comparison figures.
How do fermentation time and drying temperature change the volatile compounds in cocoa?
How
Beans were fermented for 2, 4, 6 or 8 days and dried at 60, 70 or 80 °C, and compared with sun-dried and dryer-dried controls; 58 volatile compounds were identified.
Finding
Six days of fermentation was enough to produce the volatiles associated with desirable notes while avoiding those associated with off-flavours. Drying at 70 or 80 °C after six days gave a volatile profile similar to sun drying.
Does not establish: That six days suits other cacao. Elsewhere in this register, another genotype developed its characteristic volatiles after two days and gained nothing after four.
Limits: One set of beans in one place, judged by volatile compounds in the beans, not by tasting chocolate. The 'optimum' is for those beans.
Can aroma-producing and pectin-degrading yeasts, added as a starter, change the flavour of the chocolate?
How
Small tray fermentations were inoculated with mixed starters containing Pichia kluyveri or Kluyveromyces marxianus, sampled over 120 hours, and compared with a spontaneous control; un-conched chocolates were assessed by a sensory panel.
Finding
The added yeasts made up only 9.3% and 13.5% of the yeast population at 12 hours and were never dominant. The chocolates from both inoculated fermentations were nonetheless judged to differ in flavour profile from the control.
Does not establish: That the difference is an improvement, or that it would survive conching and scale.
Limits: Small-scale trays, one trial of each, and un-conched chocolate. The abstract says the profiles differed, not how.
Are lactic acid bacteria necessary for cocoa bean fermentation?
How
Controlled laboratory fermentations in which lactic acid bacteria were restricted, compared with fermentations by the beans' own microbes.
Finding
The beans fermented fully and gave acceptable chocolates with no differences in sensory rankings. The authors conclude that lactic acid bacteria may not be necessary.
Does not establish: The same at farm scale. A 2020 review by other researchers (De Vuyst and Leroy, FEMS Microbiology Reviews) still gives these bacteria several roles in the process, including feeding the acetic acid bacteria.
Limits: Laboratory scale. The abstract does not give the mass fermented or the number of repeats.
When during fermentation and sun drying are cocoa's polyphenols lost?
How
Nacional cocoa was fermented for six days in the field; each day a sample was sun-dried and its polyphenol content and antioxidant capacity measured.
Finding
Polyphenol content and antioxidant capacity fell significantly during the first two days of fermentation. Changes over the following days were smaller. Epicatechin followed the same course.
Does not establish: How much is lost in total, or that a shorter fermentation gives better cocoa. The authors' aim was high-polyphenol cocoa, which is not the same goal as flavour.
Limits: One variety and one field trial. The abstract reports the pattern, not the size of the loss.
What do lactic acid bacteria and acetic acid bacteria each contribute?
How
Aseptic small-scale fermentations inoculated with yeasts alone, or with yeasts plus one or both groups of bacteria, and sensory tests on the chocolate.
Finding
Sensory tests showed no significant differences between the chocolates.
Does not establish: That bacteria make no difference to flavour under farm conditions.
Limits: Small aseptic fermentations with chosen strains, which is a long way from a heap.
What does cocoa pulp smell of, and do its odorants pass into the bean during fermentation?
How
The odour-active compounds of pulp from three varieties were analysed, and selected aroma compounds were measured in pulp and bean at several points during fermentation.
Finding
Pulp of CCN-51 had low flavour intensities; pulp of two other varieties had stronger floral and fruity notes. During fermentation, 2-phenylethanol and 3-methylbutyl acetate were much higher in the pulp than in the bean, while linalool and 2-methoxyphenol were higher in the bean.
Does not establish: That a fruity pulp makes a fruity chocolate. The study did not follow the compounds through drying and roasting.
Limits: Three varieties, selected compounds. The authors call it a basis for further investigation.
Which post-harvest choices most change the microbes in a fermentation and the volatile compounds in the beans?
How
Beans were fermented in wooden boxes in Côte d'Ivoire for different lengths of time, with and without turning, from pods stored for two different periods; microbial communities and volatiles were profiled.
Finding
Fermentation time had the largest effect on the microbial community, then turning; pod storage had a minor one. The same order held for the volatile profile, except that pod storage did significantly affect the volatiles.
Does not establish: What duration or turning schedule is best. The study ranks influences; it does not recommend a protocol.
Limits: One location. Volatiles in beans were measured, not the flavour of chocolate.
What do ten years of studies say about the microbes of cocoa fermentation and the starter cultures tried?
How
A review of 47 studies published in English from 2008 to 2018, from 13 countries.
Finding
The wooden box was the commonest method studied, then the heap. Thirty-seven per cent of the studies did not state the cocoa variety. Most identified microbes by fingerprinting methods and few used sequencing that could show what the microbes were doing. Yeasts were more diverse than bacteria.
Does not establish: Any effect of a starter culture. Its value here is in showing what the underlying studies cannot support: comparisons between varieties, when a third did not record the variety.
Limits: A narrative count of the literature in one language over one decade.
How do the bitter compounds, polyphenols, volatiles and flavour of one genotype change over a deliberately prolonged fermentation?
How
Beans of the Sul 1 genotype were fermented for ten days, with composition and sensory profile followed throughout.
Finding
The first 48 hours largely determined methylxanthine and polyphenol content, and fermenting beyond 96 hours reduced them no further. The volatiles characteristic of cocoa developed mostly after 48 hours. Seventy-two hours kept considerable polyphenols while giving cocoa mass with a good sensory profile.
Does not establish: That 72 hours is right for other genotypes. It is one of the few studies that names its genotype at all.
Limits: One genotype, and the abstract does not describe the fermentation's size or method.
Does the method of fermentation change which bacteria and fungi are present?
How
Bacterial and fungal communities in box, ground and jute fermentations were compared by sequencing.
Finding
Box fermentation had the higher diversity of bacterial species and ground fermentation the wider fungal community. One lactic acid bacterium and one yeast were found in all three. Pseudomonas fluorescens was abundant in the ground-fermented samples. The authors judged the box method preferable on the microbes present.
Does not establish: That box fermentation makes better cocoa. A systematic review in this register found no specific relationship between microbes and fermentation method across studies.
Limits: The abstract does not give the number of fermentations or the country. The preference rests on which microbes were found; nothing was tasted.
Can part of the pulp be taken off before fermentation, for other uses, without spoiling the cocoa?
How
Twelve variations in Ghana over three runs: with and without mechanical depulping, one or three days of pod storage, and three to seven days of fermentation, assessed by cut test, fermentation index, key odorants and tastants, and a sensory panel on the liquors.
Finding
Removing 17–20% of the pulp by weight did not harm bean quality. After five to six days with one day of pod storage, 49% of beans were well fermented with depulping and 48% without. Liquors from depulped beans had significantly less volatile acid and fewer esters, with no negative effect on the sensory profile.
Does not establish: That more pulp could be removed, or that the result holds for cacao with less pulp to begin with.
Limits: Three runs in one country. Under half the beans were classed as well fermented in either case.
Can small quantities of cocoa be fermented in a laboratory so that the flavour of a single genotype can be profiled quickly and reliably?
How
A method of fermenting small quantities of cocoa in a thermostatically controlled incubator was developed at a national cocoa research institute and compared with commercial fermentation. The abstract gives no quantities, number of runs or panel details.
Finding
The authors report that the method reproduces the basic physical and chemical attributes and microbiology of a commercial fermentation, and that statistical analysis indicated no significant difference between cocoa flavour attributes generated by the two methods.
Does not establish: No significant difference in one comparison is not proof that the methods are equivalent, and nothing here shows the method working outside the institute that devised it.
Limits: The abstract is brief and gives no numbers: no sample size, no batch mass, no statistics beyond the statement of no significant difference. Published in a national journal by the method's developers.
Whether lactic acid bacteria matter at farm scale, where the laboratory result has not been repeated.
Whether a starter culture improves flavour reliably enough to be worth its cost. On farms it has been shown to take over the fermentation and to change the flavour; no study read here shows it beating a well-run spontaneous fermentation in a blind comparison.
How long to ferment. One study found six days right and eight too many; another, on a different genotype, found most of the change complete in two days and nothing gained after four. A third of the studies in one review did not record what cacao they used.
Why Criollo is said to need two or three days and Forastero five to seven. No study comparing them side by side was found; the figures appear to be convention.
How much flavanol is lost in fermentation. Published figures range from under a fifth to more than four-fifths.
Whether heap, box or tray is better. Method changes which microbes grow; no study read here ties that to the chocolate.
These are questions the studies read for this page leave unanswered. They are not a claim that nobody has answered them.
We would like to set two Google Analytics cookies to count page views, and to count them in our own cookie-free measurement as well. They do not identify you and nothing is shared with advertisers. Decline and the site works exactly the same — see cookies for what each one does.