Hungry Ghost Food Tasting Experiment

"In this experiment, we found no difference in favor between the offered and un-offered food to Hungry Ghosts and thereby debunked the superstition"

 
Two sensory evaluation experiments were conducted, one formal by using triangle test on the cheesecake samples,
the other informal one is to try the two tables of food one offered and the other one un-offered to hungry ghosts

Written by SPI Sunkist (SPI Department of Historical and Clinical Research)

¡@

Background

How do the underworld ghost receive their burnt offerings? It is said that offerings like paper houses, paper computers, clothing, paper maids that hungry ghosts will receive by ¡§smelling¡¨. It is said that the receivers considered received the offering by the process of ¡§smelling¡¨.

Same goes for physical food offerings, the ghosts will considered ¡§eaten¡¨ the food and drinks offerings by the process of ¡§smelling¡¨. Therefore general belief that during the ¡§smelling¡¨ of the physical food, the ghosts will take away the smell and taste of food, thus leaving the food bland, tasteless and rots faster than usual.

On August 15th, SPI took the opportunity to conduct an experiment to test whether this superstitious belief is true or not. Certified Taoist priest Wei Yi performed a 7th month hungry ghost Hell-gate breaking ritual and invited the ghosts to enjoy a feast right in front of Haw Par Villa¡¦s ten court of hell.

 
Setting up the panel table for conducting the sensory evaluation experiment

Introduction to Sensory Evaluation

Sensory evaluation - A scientific discipline used to evoke, measure, analyze and interpret reactions to those characteristics of foods and materials as they are perceived by the senses. Sensory evaluation was one of the earliest methods of quality control and it is still widely used in industry. However, the level of application depends on the situation (e.g. beer and wine tasting to operators sampling of products from production line).

Sensory Evaluation is a multidisciplinary science that uses panelists and their senses of sight, smell, taste and the hearing to measure the sensory characteristics of acceptability of food products as well as many other materials. In our case here we are testing the difference in favor between two food samples. Only favor was considered in our experiment.

The sensory properties of foods are related to three major attributes:
• Appearance - colour, size, shape;
• Favor - odor and taste;
• Texture - mouth feel, viscosity and hearing.

A discrimination test is called for when the objective is to determine whether any difference is perceived between two products. The nature of the difference is usually not specified -- it is up to the test participants to see if they can find a point of difference. Since a finding of no difference may have important business implications, failure to reject the null is an actionable outcome in these tests. Thus the power and sensitivity of the test is important, and beta risk is an important consideration.

If the difference is studied as a function of different levels (systematically varied) of some ingredient, the experiment resembles a measurement of difference thresholds. For example, the determination of a just-noticeable difference is closely analogous to the discrimination test objective, when several products with different levels of a flavor are compared to some control. This is logically related to historical psychophysical methods such as the constant stimulus method.

 
SPI members were invited to participate in the sensory evaluation; Everyone was given a set of 3 samples marked in different symbols

 
This is the correct step of doing the test: Rinse the mouth before eat, and then quietly evaluate the favor

 
The panelists, Icestormer and Ironwolf were trying hard to differentiate the favors of the cheesecake

What is triangle test?

Triangle test is a type of Difference test to determine if there is a sensory difference between two products. It is used to determined an unspecified sensory difference between two treatments. The difference can involve one or several sensory attributes, but no direction or magnitude of the difference is measured.

Its principle is this: Three samples, two of which are identical, are presented simultaneously to each panelist for testing in a predetermined order. The panelist is told that two samples are identical and one is different (odd). The panelist is required to identify the different sample. The triangle test is a forced choice test.


Objectives

To identify any significant differences in taste between offerings food and un-offered food.
To identify if cheesecakes offered to ghosts become bland and tasteless.


Methodology

Triangle test (Blind Test) was conducted with 11 volunteered SPI members and one elite member. Each panelist was given one set of samples. Each set had three samples, two were un-offered cheesecake and one was offered cheesecake.

Min. Panel size of 24 people is recommended for accurate readings but only 12 people volunteered.

Each panelist was asked to taste the samples from left to right to select the odd sample out which they think tasted different. Six orders of presentations were utilized and randomly presented to each panelist. Potable water was provided to each panelist for rinsing mouth before tasting and before re-tasting. Panelists were advised not to discuss as not to influence peoples¡¦ decisions.

Instructions for panelists

• Avoid eating, drinking, smoking or chewing gum (unlikely in Singapore) for 30 minutes before testing.
• Do not talk or distract other panelists while testing.
• Read any instructions on the scoresheet before starting to evaluate samples.
• Make sure you evaluate the samples in the required order.
• Do not discuss samples with other panelists until after they have evaluated the samples.
• Have confidence in your own judgment.
• Ignore your personal likes and dislikes.

When we measure something in food (e.g. salt level in cheese) we find there is variation in what we are measuring. This variation is called natural variation or experimental error and implies that there is some true measurement but because of our limitations we cannot reproduce the correct readings every time. This is a fact of life and we have limited control over this sort of error.

Because of this variation there is some risk in making decisions about changing formulations or introducing new products onto the market. Using statistics we have rules to estimate and minimize the risk and enable us to extrapolate our results from an experiment to a more general situation.

More details can be referred to a standard namely ISO 4120:2004 Sensory analysis - Methodology - Triangle test.

 
SPI members happily volunteered in the other food tasting test that was simple and without needing strict control;
That is just to eat the food that got offered to hungry ghosts before. They commented that there were no difference in taste

 
The food although was cold and offered to hungry ghosts before, they were very delicious still

 
One strange phenomenon however, is that they seemed to be exceptional HUNGRY
They just had eaten very full about an hour ago. But they were very hungry still that is not logical

Preparation and Procedure

The samples should be representative of the product and all prepared in exactly the same way. Select four 3-digit random number codes, two for each product. Prepare scoresheets to provide equal numbers of the following orders:

AAB  BBA
ABA  BAB
BAA  ABB

Make up sets of 3 samples to match the score sheets so that half contain 2 samples of product A and half contain 2 samples of product B (Total number of sets should be a multiple of 6.). Make up sets in multiples of the six arrangements as required for the number of panelists. If total number of panelists or quantity of products available is insufficient to provide equal numbers of the 6 orders, you still need to make sure there is a balance between sets with 2 ¡¥A¡¦s and 2 ¡¥B¡¦s. The triangle tests should be presented at random to the panelists. Instruct each panelist to examine in the specified order (e.g. left to right) and remind them that they must make a decision. Count the number of correct responses (those that select the odd sample) and compare the result with those presented in a standard statistic table.

The total number of correct responses is counted as well as the total number of responses and compared to the statistical tables. This is based on the probability that if there is no real difference the odd sample will be chosen a third of the time.

A typical scenario of how the Triangle Test was conducted in laboratory

Calculate the discrimination threshold

The triangle test is used when we want to know if there is a detectable difference between two samples or products. Three samples are presented where two are the same and one is different. Panelists are asked to pick the odd one out. Purely by luck the panelist has a one in three chance of getting it right.

This forms the basis of the normal approximation to the binomial test. Lets accept that:



where
p_obs is the proportion who answered correctly i.e. X/N
p_exp is the proportion of people who we expect by chance is 1/3
q = 1 - p_exp
z is obtained from tables and for a one tailed risk of 5% is equal to 1.65.

By substituting into the equation and solving for X we get


and X = 0.778¡Ô N + N / 3

Now for a range of N values (i.e. number who sit the test) we can get a range of X values (i.e. the minimum number who must get the test right). These values have conveniently been calculated and are already tabulated for use (See table 1).

In our experiment, we have N = 12 panelists we must have at least

0.778¡Ô 12 + 12 / 3
or 6.6951 correct to achieve significance.

Since we cannot have 0.6951 of a person so we round up to 7. Therefore 7 out of 12 people must get the triangle test right in order to reject the null hypothesis and conclude there is a difference among the samples.

 
1 & 2. Sunkist was counting the scoresheets and calculating the results;
3. The result was announced. There proves no difference in taste by confidence level of 95%
4. Abductboy was sharing the results with the reporters and TCS5 staff

Evaluation

Using Null Hypothesis base on 12 panelists. (Discrimination testing)

Total Number of panelists = 12
Number of correct answers = 3 (measured result)

Probability of 3 out of 12 correct responses based on H₀ = 0.819 at 95% confident level. (refer to Table 1.)

H₀ = there is no difference
H_a = there is a significant difference between the samples

Rejection Rule:

If probability p ≤ 0.05, reject H₀

Therefore the conclusion: p ≤ 0.05 because 0.819 is greater than 0.05

Hence H₀ is accepted. There is no different between the offered and un-offered cheese cake in this experiment at 95% confident level.

Minimum number of correct answers to establish significant differences = 7 panelists out of 12 panelists as calculated from above. In this case, only 3 panelists out of 12 panelists got correct answers and none of the 3 panelists commented that it is bland less.


Conclusion

There were NO significant differences between un-offered cheese cake and the offered cheesecakes during the Haw Paw Villa food testing experiment.


Possible experimental errors

1. Due to un foreseen circumstances, dinner for members was pushed from 8pm till about 9pm. Panelists might be too full to detect small differences between the offered and un-offered cheesecakes. Dinner ended at 10.15pm and food tasting begins an hour later.
2. Fatigue could be a possible reason, after the hell-gate breaking, followed by tour around 10 courts of hell, it was near mid night. Panelists might be too tired to concentrate.
3. Disturbances during sampling, there were a lot of noises and lightings not bright enough due to Haw Par Villa night environment
4. Sampling size too small, only total of 12 panelists volunteered to do the experiment. Again, fatigue and a full dinner properly the reason for the small sampling size plus the possibly of hungry ghosts following home after eating offering food.
5. As members have to rush back to take the last bus and train, food offered to ghosts might not be long enough. It was expected to leave the offerings for at least an hour.

In the future, similar food tasting experiments will be conducted by SPI again. We would like to invite you to repeat the experiments. Please send us an email if you are willing to be our panelist in our future experiments.


Selected Bibliography

• Gacula, M C and Singh, J, ¡§Statistical Methods in Food and Consumer Research, New York: Academic Press, 1984.
• Jellinek, G, ¡§Sensory Evaluation of Food: Theory and Practice¡¨, Chichester: Ellis Horwood; 1985.
• Lawless, H T & Heymann, H, ¡§Sensory Evaluation of Food: Principles and Practices¡¨, Chapman & Hall, New York, 1998.
• Meilgaard, M, Civille, G V and Carr, B T, ¡§Sensory Evaluation Techniques: Boca Raton, Fla: CRC Press, 1999. (3rd Edition)
• Miflora Minoza-Gatchalian, ¡§Sensory Evaluation Methods with Statistical Analysis (for Research Product Development and Quality Control)¡¨. 1981.
• Poste, L M, Mackie, D A, Butter, G and Larmond, E, ¡§Laboratory Methods for Sensory Analysis of Food¡¨, Agriculture Canada Publication 1864/E, 1991.

Table 1.



¡@