6.2. Sensory Analysis

After the past fifty years sensory analysis of foods has grown from a informal ‘taste test’ performed by bench top chemists and product developers to a science comprising basic tenets, accepted methods, and defined statistical analyses. The scientific literature of sensory analysis comprises investigations of sensory processes (the way we respond to physical stimuli), methodological studies of ‘taste testing’ (how to assess reactions to food products in a test situation), and statistical experimental design (how to set up systematically carried formulations of product).

The actual physiology and psychophysics of the sensory systems involved in the sensory analysis (viz, taste, smell, touch, and appearance) are discussed in Trends in Food Science and Technology, Vol.7, Nº12 Dec. 1996. Data about the physical characteristics of food belong to the domain of food science. Sensory analysis is the nexus where these different disciplines join.

• New product development
• Cost reduction exercises
• Quality improvement
• Evaluation of product acceptance
• Quality control and assurance.

• Establish and characterize changes which may be either natural to or induced in the product
• Distinguish between different samples
• Ascertain whether some defined quality attributes can be expressed in terms of a simple numerical scale or index
• Whether the total activity of the sample can be expressed in a simple or multidimensional manner so as to rank different samples
• Establish standards and specifications
• Grade samples with on without references to a standard sample
• Establish a relationship between objective and subjective test data
• Establish the hedonic values of the product.

• Difference testing - usually carried out by a small panel of assessors (judges) not necessarily having great experience
• Rating into categories - usually carried out by a panel of well-trained and experienced assessors
• Acceptability appraisal - may be carried out by panels from small local groups to vary large consumer panels. These, preferably, should not include trained technique staff whose judgments may be unduly based and unrepresentative of the majority.

• Descriptive analysis. The language of food and drink, and how we describe our perceptions.
• Intensity measurement. Perception of the sensory characteristics of foods on a scale, which shows how we process physical stimulus magnitudes to generate sensory responses.
• Hedonic measurement. Our likes/dislikes.
• GC sniffing analysis.

Descriptive analysis

At the heart of sensory analysis lies the complex field of language or descriptive analysis. While at first blush it may seem easy to describe a food, individuals familiar with a specific food use many terms to describe the fine nuances. Some of these terms are idiosyncratic, but many are standard terms easily understood. There is no single descriptive language of general and universal application, but rather attempts to capture on paper and numerically the elusive character of different foods.

Since ancient times researchers have tried to create standardized lists of terms by which to summarize (and classify) sensory perceptions. Researchers using any of these lists soon find that the set of terms in any list is unbalanced and necessarily incomplete, comprising too many unusable terms for the specific product being studied, but too few relevant terms to capture the key ‘note’ or attributes. Human beings use as stretchable ‘rubber bag’ of terms to describe each product. The number of terms grows or shrinks to fit the specific product, but the list for any product may be sure to contain terms that would never be found in a general list.

People process a constant amount of ‘information’ when they describe products. When forced to describe a wide range of qualitatively different products during a single test panellists overlook or jettison the nuances and focus on the more general terms that differentiate this broad array of qualitatively different stimuli. For products which vary only slightly from each other (e.g., different samples of strawberry aroma) panellists employ the more rarely used terms, focusing on the minor differences and highlighting them with these terms. Examples of attributes for a variety of Odorants are:

Examples of lists of terms used in the description of Vanilla extract are:

Flavour characteristics of pure Vanilla extract (these descriptions reflect the norm, flavour quality does vary between lots.)

Measurement of perceived intensity

Intensity measurement (quantification) comprises the second major application of sensory analysis. Quantification takes many forms, including determination of threshold, assessment of discrimination ability (the smallest physical change between two samples that is just noticeable), and scaling the suprathreshold intensity (how strong does the stimulus seem).

Hedonics - measurement of liking

Acceptance measurement, the third branch of sensory analysis, may well be the most vital aspect. We select and eat foods on the basis of their visual appeal and palatability. A large and growing body of scientific literature has been published on the measurement of acceptance. Appropriate measures of liking or purchase intent, and the assessment of possible consumer boredom with a product and critical issues when applying hedonic tests in the world or commercial research.

GC sniffing analysis

Aromatic extracts of foods are composed of more than one hundred volatile compounds, but the profiles obtained after GC analysis with electrochemical detectors do not necessarily reflect their aromatic specificities. Some of the compounds, present in large quantities, are not relevant to the overall aroma. Others, with low detection thresholds, are detected after GC sniffing, whereas electrochemical detectors fail to give any response. For these is defined the Aroma value that is the ratio between the concentration of a compound and its threshold.

These compounds are responsible for the aromatic characteristics of the food. As a result, sensory methods are needed to allow identification of the main aromatically active compounds and have investigated by different authors.

There are two methodologies to do GC sniffing analysis: replace the electrochemical detector with a trained sensory panel that smells the aroma-active region directly from the GC effluent (aromagram); or sniffing the aroma fractions collected from the effluent. The second method more complicated but it allows the identification of fragrances constituted by two ore more compounds.

Interpretation of results

The results of many of the descriptive and affective sensory methods described can be studied by tabling the data, including the score of each judge for each sample, the means, the ranges, and the deviations from the mean. Some of the variability in results is attributable to the samples themselves, and may be a combination of differences in the raw materials and in the method of preparation. Sources of error in the judging include variability in the performance of one judge on duplicate samples as well as variability among several judges on the same sample.

After the data have been tabulated and averaged, the answer to the question posed by the experimenter may be obvious and further analysis unnecessary. A study of the data such as described above is not adequate when the investigator wishes to state with confidence that the results obtained are statistically significant. In this case, a statistical analysis of the results is necessary. The original experiment should have been planned with statistical analysis in mind, because it is difficult, and sometimes impossible, to apply statistical to a completed experiment not appropriately planned. References have been given to tables based on statistical analysis for several discrimination and descriptive tests. Various methods of testing the significance of differences among means may be used depending on the experimental plan. Analysis of variance, with a multiple range test when appropriate, is very useful in many cases. Correlation, or an indication of the relationship between two variables, can be calculated between descriptive sensory data and objective data. Access to data handling and statistical analysis software for personal computers or to mainframe computers may facilitate data handling and analysis. The development of software for direct computer entry of data by panellists and the subsequent analysis by personal or mainframe computer will facilitate this process.

Experimental Design

Sensory analysis has achieved its major commercial impact by providing reliable test results from panellists. When coupled with experimental design and modelling (by regression analysis), sensory analysis becomes a powerful tool to guide product development.

An alternative approach beyond directional scales to develop products using consumers, consists of systematically varying the formula ingredients in a way which allows one to assess the effects of each ingredient, and the interactions among ingredients on acceptance and attribute perceptions.

Experimental designs are properly the province of statisticians. Sensory analysts and product developers have used experimental designs with great success to understand consumer reactions to test prototypes comprising known ingredients and processes. Product categories amenable to these designs range from simple food systems such as a fruit-flavoured beverage, to complex systems such as pizza, apple pie, and sausage. In all cases the assessment of systematically varied alternatives has educated the researcher and provided concrete direction for product modification.

Using the Product Model to Assure Quality

The product model relates independent attributes (under research or production control) to dependent variables, be these consumer ratings, expert panel ratings, etc. The product model plays a role in assuring quality, because it can be used in two distinct ways to control production:

1. Process Control. At the most basic level sensitivity analysis reveals how process or ingredient variations affect liking. Once the desired formulation has been located within the grid of alternative formulations, and the range of independent variables set (high to low for each independent variable), the investigator computes the sensitivity curves for each independent variable, holding the other independent variables fixed (at the prescribed level for production). The changes in the sensory ratings show how the independent variable affects the sensory character of the product (viz, sensory attributes). The change in liking indicates how changes in the independent variable affect acceptance.

   Quality control tables based upon sensitivity analyses highlight those independent variables, which produce noticeable sensory changes, and the degree to which those changes generate acceptance changes in their wake. The manufacturer should maintain tighter control over the key critical variables (perhaps at greater cost) and maintain looser control (at lower cost) over the less important variables (where departure from the optimal or production specifications do not reduce acceptance nor affect sensory integrity).

   The approach is consumer driven, because it is based upon the reactions of consumers to actual variations of the product, rather than hypothesis of what production variables might be important. It might well turn out that some variables play little or no role in determining acceptance, and can either be reduced to save money, or ignored by quality assurance over a wide range of levels. Conversely, more attention would then be paid to those important variables which influence acceptance.

2. Batch Analysis. The inter-relation among formula variables, consumer responses, instrumental variables, expert panel responses, and quality control panel responses assures quality at the production level on a batch-to-batch basis. Experts measure each batch, either, by the quality control panel, or by instruments. The measurements generate a goal profile. The goal profile determines a corresponding set of estimated ingredients or process variables, which would generate that profile. Once the levels of the independent variables are estimated the model estimates the corresponding consumer sensory profile and / or the likely acceptance rating. As each batch emerges from the production line quality assurance can calculate the expected difference from the reference or “gold standard” on a sensory or acceptance basis. At the time of the measurement, plant personnel decide whether the product is sufficiently similar to a target standard (or reasonably acceptable) to warrant shipment, or whether the batch must be reworked (or even discarded).