Introduction of a Holistic Approach to Evaluate and Improve the Sensory Attributes of Milk Products
Updated: Oct 19
Thomas Eidenberger PhD, Jack (Jingang) Shi
Abstract: This article introduced the history, concepts and sensory sequence of milk and dairy products' sensory performance analysis. It also explained in detail how to use a holistic approach to evaluate the sensory performance of milk and yogurt. The control experiment clarified that the flavor and taste of milk-containing beverages and dairy products can be optimally improved by adding specific taste modulators (flavors), thereby increasing the consumer liking of milk-containing beverages and dairy products.
Key words: Milk, flavored dairy products, sensory attributes, flavor, mouthfeel, taste modulators, consumer liking
1. Historical Perspective of Milk Sensory Analysis
Sensory evaluation of milk and milk products has an outstanding long tradition in Europe. Milk is one of the most important basic food and was traditionally consumed without further processing. It was therefore always essential to assess the quality of milk to prevent health risks in connection with consuming milk. It should be emphasized that milk is an important part of the nutrition for babies and toddlers, a group of the population for which special caution is required.
After the discovery by Louis Pasteur (1822-1895) and Robert Koch (1843-1910) that microbial contamination is a source for product deterioration and pathological reactions when consumed, the safety of milk and milk products increased substantially. Raw, unprocessed milk may can contain Salmonella Serovar Typhimurium, entero-hemorrhagic E.Coli (EHEC), Campylobacter spp. or Listeria spp. Raw milk was the main source for transfer of tuberculosis in the past. After introduction of hygiene standards for the milking of cows (sheep and goats) and heat treatment milk has become a safe consumer product and a safe raw material for the Dairy Industry.
With increasing processing of raw milk before consumption the targets for sensory evaluation shifted from qualitative, subjective assessment (i.e. good-bad, fresh-old) step by step to the objective assessment of sensory properties and liking testing to investigate consumer preferences. As early as 1920-1930 the first score cards for milk assessment have been introduced (Nelson, J. A., and G. M. Trout. 1934. Judging market milk. Pages 31-49 in Judging Dairy Products. Olsen, Milwaukee, WI.).
2. The Concepts of (Milk) Sensory Analysis
The simplest concept of sensory analysis is the defect judgment. The only target is to assess whether milk shows defects while tasting. Typical taste/smell defects are
cooked (caused by improper heat treatment)
oxidized (caused by light, temperature during storage and presence of metals)
feedy (caused by transfer of feed ingredients into the milk)
sour (caused by fermentation)
It should be emphasized that defect judgment is a qualitative Yes/No decision and neglects any intensity or more detailed descriptive rating.
As of today the standard concept for milk sensory analysis is the descriptive sensory analysis.
To perform descriptive sensory analysis adequately, a panel of 5-15 trained tasters is required. The target of the training is the reproducible recognition of sensory attributes including their intensity. Moreover the sensory conclusion should be similar between the tasters. To achieve reproducible and comparable conclusions, a set of sensory attributes has to be fixed and reference preparations/ products with different intensities of each sensory attribute chosen have to be defined. Find on following Table 1 examples for a milk taster’s training schedule.
Table 1, Examples of Sensory Attributes and the Reference products to assess their intensity
(Adopted from Schiano AN, Harwood WS, Drake A. A 100-Year Review: Sensory analysis of milk. (2017), J. Dairy Sci. 100: 9966-86).
The evaluation of descriptive sensory analysis includes statistical methods to yield objective results and follows a pre-fixed decision tree to secure reporting of adequately performed sensory tests only.
Usually the same testing, assessment and evaluation procedure is applied to various samples of the same product category to investigate differences between the samples tested.
The sensory concepts described above do not provide evidence for preferences between samples or ranked liking. If major defects are observed, it would certainly correlate with preference and liking. If neutral sensory attributes of milk products (sweet, creamy) are tested the test results will not correlate to preference/liking. Exempt are test designs where a “golden reference” is defined and the difference to the golden reference is indirectly related to preference.
The performance of (quantitative) descriptive sensory analysis including selection and training of tasters, the methodologies and design of tasting rooms is regulated in a series of ISO-Norms (searchable at www.iso.org).
Consumer preference testing (hedonistic tests) consists of test designs to evaluate the overall liking of products. The most simple version is to distribute one sample to a group of randomly selected intended consumers (eventually restricted to gender, age, ethnicity or other demographic characteristics) and to ask them to give a rating (i.e. 0 (total rejection) to 5 (brilliant)) after tasting. Such tests can be extended to several samples and comparative ratings. If not performed solely for marketing reasons, consumers can be invited into a sensory test facility and are simply trained by explained tasting of reference samples before blinded tasting of several samples is started.
In contrast to descriptive sensory analysis, hedonistic preference tests yield a quantitative measure for consumer preference (i.e. product A is highest rated by 70 % of consumers) but does not provide any information on sensory attributes triggering a liking decision.
3. Sequence of Sensations during Tasting
To ultimately link sensory analysis with liking it is necessary to understand the timed sequence of sensations when tasting food.
The first sensations are the color, the visible texture and the temperature. That alone can already trigger a sensory decision. It is well known that butter with a yellow/orange color is rated softer than butter with a white color without tasting and irrespectively of the temperature. Yoghurt with a watery supernatant will not be rated brilliant irrespectively of the smell and taste.
In case of unblinded tasting (i.e. yoghurt in the branded cup) our brain may also trigger a decision solely on basis of the brand and/or the design of the cup.
The second sensations are related to the smell -more precisely the ortho-nasal sensations while breathing in- of a product. The smell or in case of pleasantness “aroma” is a very important decision-making part of sensations as it determines whether we ingest food or not. Interestingly the recognition of a smell in our brain is more important than the pleasantness. Ripened cheese which sometimes possess a strong and rather awful flavor is recognized as having perfect taste and texture. Disturbing for a decision is an unexpected or unknown smell.
When ingesting food, mastication and contact with the tongue, the palate, the cheek’s mucosa creates a variety of sensations which bombard our brain. The duration of mastication and the movement of the food in the mouth can influence the overall sensations and also change them over time (degradation of starch to glucose due to enzymes in the mouth causing sweet taste).
Constant breathing while chewing prompts retro-nasal sensations while breathing-out. Retro-nasal sensations affect the overall smell but also taste perception. Evolutionary, the ortho- and retro-nasal “detection-system” are one of the most well preserved systems and functioned anatomically very similar in Dinosaurs and humans. One explanation is that ortho- and retro-nasal sensations are extremely important to prevent animals and humans swallowing of harmful food. While ortho-nasal sensations are acting as first barrier preventing ingestion, retro-nasal sensations act as last barrier preventing swallowing.
Swallowing of food triggers anatomically opening of the retro-nasal cavity which then flooded with volatiles in the food by breathing-out immediately after swallowing. Further on flavor receptors in the throat induce after-taste sensations.
4. Correlation of Sensory Attributes with Liking
From above explanations it becomes clear that any accurate description of sensory perception of food does not only require assessment of sensory attributes with intensity rating but also assessment of sensory attributes over time.
To perform such tests, the change of individual attribute intensities over time needs to be recorded. Finally, individual time/intensity curves are combined to assess the different sensations and their timed sequence. Find on following Figure 1 two diagrams showing sweetness, acidity and flavor intensity over time in vanilla yoghurt without and with 100 ppm Savarin™ 100PN (EPC Natural Products, Beijing).
Figure 1, Time/Intensity curves for sensory attributes in Vanilla Yoghurt and Vanilla Yoghurt with added Savarin™ 100PN (PN, 100 ppm), Intensity Rating: 0=none, 8=maximum
As seen in Figure 1, adding of Savarin™ 100PN to Vanilla Yoghurt changes the timed sequence of sensations substantially. The overall flavor profile with added Savarin™ 100PN is rated much better due to less lingering flavor, harmonic sweet/sour balance and an embedded vanilla flavor.
Apart from the comparable assessment of time/intensity profiles of sensory attributes, the assessment of the Temporal Dominance of Sensations (TDS) has been shown as a valuable tool to compare different products (Pineau N., et al. Temporal Dominance of Sensations: Construction of the TDS curves and comparison with time–intensity. (2009) Food Quality and Preference 20, 450–455. Instead of recording individual time/intensity curves TDS describes the dominant sensation at a given time point as number of tasters divided by all tasters (i.e. 8 out of 10 tasters rate the same sensation dominant gives a fraction ratio of 0.8). The raw data are then transferred to curves (spline interpolation) and finally the difference between 2 products can be visualized by a difference diagram as shown in Figures 2 and 3.
Figure 2 shows the comparable TDS curves for a Strawberry Yoghurt sweetened with artificial sweeteners after addition of 200 ppm Savarin™ 100PN. Significant shifts in TDS ratios are observed for acidity (front peaks disappears) sweetness and flavor (prolonged dominant sensation).
Figure 3 shows the difference of the two products obtained by subtraction TDS ratios of Strawberry Yoghurt from Strawberry Yoghurt with Savarin™ 100PN. Positive differences stand for more dominance in Strawberry Yoghurt with Savarin™ 100PN while negative values stand for less dominant. Strawberry Yoghurt sweetened with artificial sweeteners and added Savarin™ 100PN is at the beginning significantly less acidic, followed by increased and prolonged sweetness and prolonged flavor perception. The differences are correlated with a much better overall rating for the Strawberry Yoghurt with Savarin™ 100PN (more sugar like sweetness, balanced sweet/acidic balance, more natural flavor perception).
Figure 2, Temporal Dominance of Sensations (TDS) in Strawberry Yoghurt Sugar-free (SF, Acesulfame K, Sucralose) and Strawberry Yoghurt with added Savarin™ 100PN (PN, 200 ppm)
Figure 3, Difference in Temporal Dominance of Sensations (TDS) in Strawberry Yoghurt with added Savarin™ 100PN (PN, 200 ppm) and Strawberry Yoghurt Sugarfree (SF, Acesulfame K, Sucralose)
5. Texture and Sensory Perception
The texture of food plays a significant role for the overall sensory perception. Crispiness, hardness, softness, viscosity after melting in the mouth, warming/cooling effects are important descriptors for (semi)solid food. Liquid or creamy food is usually rated by the mouthfeel. Watery, void, flat are typical mouthfeel defects while full-bodied, rich, smooth are used to describe a pleasant mouthfeel. In addition to the texture mouthfeel is influenced by refreshing/ripe and contracting/ coating effects.
It should be emphasized that the fit of mouthfeel for an individual product depends very much on the expectation and the experience of the taster. French soft cheese (i.e. Camembert) is expected to be almost gelatinous, Greek yoghurt (10 % fat in dry mass) is expected to be very creamy. If such expectations are not met (i.e. unripe Camembert with solid, curd-like texture, Greek Yoghurt with watery texture) rating of mouthfeel is low irrespectively of whether objectively mouthfeel is good or bad (i.e. if tasted blinded a slightly watery Greek Yoghurt will be rated good in mouthfeel as the expectations for regular Yoghurt are different from “Greek Yoghurt”).
Texture of food exerts an important influence on flavor perception as volatiles can be entrapped in the food matrix or receptors in the mouth can be masked by matrix components.
To connect the impact of mouthfeel on the overall flavor perception, more recently a “Cube Model” for universal flavor factors has been described (Klosse PR. Flavor Classification, Another Paradigm. Towards a universal flavor model, (2017), DOI: 10.13140/RG.2.2.35017.65126).
This model considers three basic aspects of mouthfeel, namely contracting, coating and drying, as well as two aspects of flavor, namely freshness and ripeness, and additionally the flavor intensity to define product categories in a three-dimensional cube model.
Contracting mouthfeel is triggered by fresh and sour taste (citrus fruits, spicy food, pungency), coating mouthfeel is triggered by ripe and full taste (fatty, creamy, sweet fruits like peach), dry mouthfeel is perceived with balanced mouthfeel (meaty, mushrooms). Depending on the flavor intensity (low-high) 8 basic flavor style can be defined as shown below in Figure 4.
Recently, to connect the impact of mouthfeel on the overall flavor perception, a model for universal flavor factors has been used (Klosse PR. Flavor Classification, Another Paradigm. Towards a universal flavor model, (2017), DOI: 10.13140/RG.2.2.35017.65126), see Table 2.
Table 3, Characteristics of basic flavor styles (adopted from Klosse PR., 2017)
Figure 4, Cube model for 8 basic flavor styles based on dominance of mouthfeel perception and flavor intensity. Exemplifying high contracting, low coating and dryness resembles freshness, Low contracting and dryness, high coating resembles ripe fruits.
In this model, beverages are classified according to the mouthfeel (contracting, coating, drying) and flavor intensity/complexity (low, high). Contracting mouthfeel is experienced as tactile "tingling" or "stinging" impression which is perceived as a refreshing feeling. Acidity, saltiness, and coldness (ice cubes) trigger a contracting mouthfeel in the mouth. Refreshing is closely associated with a dominant mouth contracting and a low flavor intensity/complexity (equivalent to quick recognition).
The cube model is a valuable tool to include mouthfeel into the overall assessment of sensory analysis. Again, the expectations before ingesting are the most important predictor for the rating. Regular milk is expected to be coating, hence low fat milk is rated low as it appears watery and flat. Long ripened hard cheese is expected to be drying and pungent, hence taste is not satisfactorily when tasted after too short storage (too soft, mild, buttery).
The cube model is especially valuable to assess these expectations as the 8 basic quadrants needs to be met for adequate liking. Spicy and contracting doesn’t go well together (vinegar & chili) whereas spicy and dry (meat & chili) or spicy and coating (sour cream & chili) fits much better.
Modification of mouthfeel is nowadays gaining more and more importance. Sugar, salt and fat reduction are becoming more and more popular in the industrialized world, the resulting product however are usually poor in flavor or need additional food additives to restore adequate mouthfeel. EPC Natural Products has developed flavors with flavor modifying properties to restore adequate mouthfeel without adding food additives.
Exemplifying Figure 5 shows the effects on mouth coating and freshness in a sugar-free Lemon Yoghurt (sweetened with Acesulfame K and Sucralose) during storage at 8 °C.
Figure 5, Perception of Mouth-Coating and freshness in Lemon Yoghurt Sugarfree (SF, Acesulam K, Sucralose) without or with added Savarin™ 100PN (PN, 200 ppm) over 2 months of storage at 8 °C.
5. Closing Remarks
The information provided in the previous chapters should provide some basic knowledge about how to investigate sensory attributes and use the results to estimate the liking of a product.
The examples and techniques are by far not comprehensive, there are many more test and evaluation designs (i.e. triangle tests to identify whether products taste different, Principal Component analysis (PCA) to cluster similarities/differences between products, and many more).
The examples and techniques described, however, provide a tool box to link flavor attributes to mouthfeel and relate it to liking.
Flavors can be crudely categorized following attribute pairs
Type “Fresh” (lemon, menthol) or “Ripe” (peach, garlic)
Intensity “weak” (milk, butter) or “Strong” (vanilla, cooked meat)
Style “Neutral/Balanced” (water/white bread) or “Pungent/Complex” (pepper/mushrooms)
These opposite terms mark cornerstones and in reality the overall perception may be a combination of both opposite terms or even between the categories.
In addition to the flavor type, intensity and style the predominant matching factor to be met for liking is the mouthfeel with is categorized crudely as
Coating (Milk, Honey)
Contracting (Vinegar, Apple)
Dry (Rice, Cacao)
High liking of a product is generated by the optimum combination of flavor type, intensity and style with the mouthfeel being predominantly coating, contracting or dry. When considering a milk drink (coating mouthfeel) adding lemon juice (fresh, strong and pungent) will create a misfit whereas adding banana (ripe, weak, balanced) creates an optimum fit with high liking.
Summarizing it can be concluded that sensory attributes help to categorize the flavor tape, intensity and style together with the category of mouthfeel. Based on these categorizations, the expected liking of product can be objectively forecasted. If not satisfying (enough) improvement of the recipe to optimize the fit and to increase liking is strongly recommended.
 Nelson, J. A. & G. M. Trout. (1934). Judging market milk. Judging Dairy Products (pp.31-49), Olsen, Milwaukee, WI.
 Schiano AN, Harwood WS, Drake A. (2017). A 100-Year Review: Sensory analysis of milk., J. Dairy Sci. 100: 9966-86
 Pineau N., et al. (2009). Temporal Dominance of Sensations: Construction of the TDS curves and comparison with time–intensity. Food Quality and Preference 20, 450–455
 Klosse PR. (2017). Flavor Classification, Another Paradigm. Towards a universal flavor model, DOI: 10.13140/RG.2.2.35017.65126