One of the most exciting and dynamic areas of sensory research is into methodology to determine perceptual changes that happen over time.

These can be as different as the fleeting changes that take place in the bouquet of a fine wine or the tedium faced by hospital patients as they eat another institutional meal. There is no doubt that the changes that take place over time are important, but they also represent a challenge.

In the 1980’s the interest in Time Intensity measurement of high-intensity sweeteners lead to a flurry of publications and the development of computerized methods for conducting single-attribute data collection. For sweetness, this was possible and a lot of information was gathered and a whole range of analyses were applied. The biggest challenge was that although we could get reasonable data collected on model systems, real products with complexity presented a much bigger problem. Trained descriptive panellists were being asked to respond to a complex stimulus using a single attribute. This leads to the classic bias of “dumping” or attribute restriction.

In response to the need to extend Time Intensity measurement, we developed Dual-Attribute Time Intensity. DATI panellists move a mouse to record two attributes simultaneously on two axes on the screen. The publications on sweetness and flavour measurement in chewing gum and on tenderness and juiciness in cooked meat proved the technique to be useful. (See the Research section for article summaries and downloadable PDFs about Temporal Methods)  But the training required to get the panellists to a comfortable and reproducible state was extensive and more than most sensory labs wanted to take on.

Time Intensity Plot

Concurrently, we were performing Descriptive Analysis on commercial beer. The trained panellists used a structured ballot that was derived from a set of Canadian and US domestic liquids. When the scope was expanded to include European and International beer, I noticed that panellists were picking their attributes in a different sequence in response to a different order of perception. We were also able to observe the paradox of beers with profiles that were not significantly different, but easy to discriminate due to temporal attribute release differences. This led to a project called Temporal Attribute Difference (TAD). Products were evaluated for intensity of a series of attributes and the time to select and score each attribute was recorded using experimental software. The research was performed on a salad dressing model system which allowed us to measure salt, sweet, sour and bitter taste and the flavour of garlic. Viscosity and oil proportion were modified to create products that demonstrated difference in the time of perception of attributes for the case of shuffle (a change in perceived order of attributes) and shift (a temporal delay of all attributes). I gave a paper at the IFT meeting in Dallas in 2000 (see “Temporal attribute discrimination” in the Compusense Research Library) with the sub-title “Why this method does not work”. Although we were able to measure both “shift” and “shuffle” in the model system, when the method was applied to real products the variance became very large. The noise in the data came from the complexity of the task. The time measured included three components; (i) the time required for sensory identification, (ii) cognitive processing to find the correct attribute on the screen and (iii) the coordination required to mark the intensity on the line scale. The method was neither accurate nor precise. The work was not published for a number of reasons. Mostly because it was a disappointment, but it was not part of a student’s thesis, so it was allowed to fade away. In retrospect, even research failures should be published to help future researchers.

The pursuit of complex temporal measurement has continued and Pascal Schlich and his students at the Centre Européen des Sciences et du Goût at the University of the Bourgogne in Dijon have been leading the charge. They have developed a method called Temporal Dominance of Sensations (TDS). TDS followed in the footsteps of TAD and attempted to measure time of selection and scale value, but only for the dominant attribute. There has been some ambiguity around the working definition of dominance. It may be the strongest attribute at a particular time or it may be the attribute that comes to the attention of the assessor. The data that has been collected using the line scale intensity have been very noisy, which we saw with TAD. So it is only the temporal information that has been gathered that is used to simulate a TDS curve based upon frequency of perception. At any time point, the attribute that is being selected by the largest number of assessors is called “dominant”. If it passes a frequency threshold it is then called significant. So the TDS curve is a mathematical construct that uses frequency to simulate intensity. This provides some qualitative data that may help product developers understand product performance. In practice, quite large numbers of assessors, over 30, are used. The numbers of attributes that can be measured are also limited, with best results obtained from five at a time. Ultimately, the TDS method is not very repeatable. (Meyners, Sensometrics 2010)

A more promising approach to temporal intensity measurement is Progressive Profiling as applied by Lisa Methven at Reading University in the UK. She uses a descriptive profiling screen that appears at predetermined time intervals. All attributes (usually 5) are seen and scored at each interval. The curves that are produced are real time curves. Not only can you determine the strongest attribute, but you can follow every attribute over the course of the experiment.  The assessor does not have to judge dominance.

It would be desirable to follow multiple attributes easily and reliably, but we are not there yet. As progress is made in measurement techniques and in statistical methods that help us handle noise, there will be many opportunities to move sensory temporal research forward. It is a noble and worthy goal; after all these dynamic measurements reflect the sensory realities of consumers.