Castura, J.C. Food Quality and Preference, 2010, 21(3), 257-258. Refereed Publication.

Equivalence testing has applications that include ingredient substitution and product matching. Statistical methods for determining equivalency were the subject of some interest in this journal prior to the Sensometrics 2008 conference (Bi, 2005; Meyners, 2007; Bi, 2007; Ennis, 2008a; Bi, 2008; Meyners, 2008; Ennis 2008b). A mini-symposium on equivalency at Sensometrics 2008 provided an opportunity for collegial discussion. Dr. D. M. Ennis presented methods for testing equivalency for binary data, normally distributed data with known variance, and normally distributed data with unknown variance. These tests are well described in recent publications (Ennis & Ennis, 2009a; 2009b). The two one-sided tests (TOST) procedure (Westlake, 1981; Schuirmann, 1987) is commonly used when testing equivalency. The TOST procedure is often used with tests that make parametric assumptions, but can be used with tests that do not make such assumptions (Zhou & Yuan, 2004). Ennis presented the adjusted noncentral chi-square (ANC) test as an alternative procedure to the TOST for normally distributed data with unknown variance. Drs. P. B. Brockhoff and M. Meyners accepted invitations to provide critical feedback on Ennis’s presentation. This review attempts to capture some of the key points.

Findlay, C.J., Castura, J.C., Lesschaeve, I. Food Quality and Preference, 2007, 18(2), 321-328. Refereed Publication.

Training targets were established using descriptive analysis profiles of 20 commercial red wines produced by a well-trained, experienced determination panel. After recruitment, screening and a basic sensory orientation of ten 2h common training sessions, 16 inexperienced panellists were divided by lottery into two panels. The control panel received a more conventional performance debriefing at the end of each training session. The experimental panel only received immediate graphical computerized feedback while in sensory booths.

Both panels evaluated the same 20 wines and used the same scales and attributes. Panels were calibrated and responses compared to training targets. Performance was monitored daily as panels continued over a three-week period. Distance from target measurements showed similar improvement trends for both groups as measured by panellist and panel calibration. Results suggest the effectiveness of the feedback calibration method (Compusense FCM®) in providing unbiased and effective training.

Download - 212 KB PDF
Categories: Analytical Sensory Methods; Compusense FCM^®; Computerized Sensory Analysis;Descriptive Analysis; Wine;

Findlay, C.J., Castura, J.C. Schlich, P., Lesschaeve, I. Food Quality and Preference, 2006, 17(3-4), 266-276. Refereed Publication.

The performance of descriptive panels is typically determined by post-hoc data analysis. Poor panel performance is measured after the fact and often arrives too late to help the panel leader during training sessions. The feedback calibration method (FCM) optimizes proficiency by ensuring efficient panel training. A previously trained panel (Panel T) and an untrained panel (Panel U) developed and refined their own training targets using FCM before evaluating 20 white wines in triplicate. Permutation tests of the RV coefficient were used to compare the panels in terms of the underlying sensory space.

The results of the panels were similar, both Panel T and U were superior to a proficient conventionally trained red wine panel (Panel D). Panel U performed similarly to Panel T on proportion of attributes discriminated and disagreement using a two-way mixed-model analysis of variance (ANOVA) and on multivariate discrimination evaluated by a MANOVA with the same mixed model. Evaluation means for product*attribute fell within the training range targets in 59% of the cases for Panel T and 69% for Panel U, providing an indication of the panels’ abilities to hit the training targets. Panel U was shown to be proficient in discriminating a full range of wine attributes (p = 0.05) after only nine formal training sessions (22.5 h), a reduction in training time of 49%.

Download - 157 KB PDF
Categories: Analytical Sensory Methods; Compusense FCM^®; Computerized Sensory Analysis;Descriptive Analysis; Wine;

Castura, J.C., Findlay, C.J., Lesschaeve, I. Food Quality and Preference, 2005, 16(8), 682-690. Refereed Publication.

Training targets can be established from product profiles that provide an objective representation of the underlying sensory characteristics of a group of products. If available to a panel leader, these training targets can be used to calibrate a panel and measure the accuracy of their responses.

Response accuracy can be determined either by frequency counts—how often the target was hit versus how many opportunities there were to hit it—or by distance from target measurements—which attempt to further quantify the degree to which the target was hit or missed.

In addition to the frequency counts, four distance-from-target methods are presented and discussed—Distance from Target, Distance from Range, Adjusted Distance from Target, and Adjusted Distance from Range—each of which provides insights into the degree to which the panel is calibrated.

Download - 145 KB PDF
Categories: Analytical Sensory Methods; Compusense FCM^®; Descriptive Analysis; Statistical Methods and Data Visualization;

Findlay, C.J. Food Quality and Preference, 2002, 13(6), 423-428. Refereed Publication.

Computer technology is changing rapidly as is the scope and use of the Internet. These tools are being applied to a broad range of quality control activities, including sensory evaluation. The main areas of impact of this technology are in test design, collection of data, tabulation, storage, statistical analysis and reporting of the data in real time over great distances. Effective quality systems can be constructed using anything from the simplest spreadsheet programs through to sophisticated integrated quality control systems operating over corporate networks. This article provides an overview of the tools that are available and discusses a specific case as an example of a starting point for computerizing sensory quality control.

Download - 716 KB PDF
Categories: Sensory Quality Control; Computerized Sensory Analysis;

Zimoch, J., Findlay, C.J. Journal of Food Science, 1998, 63(6), 940-944. Refereed Publication.

We examined the effectiveness of Dual Attribute Time Intensity (DATI) method for assessment of temporal changes in perceived toughness and juiciness, within commercially acceptable meat cuts. Usefulness of DATI in assessing temporal aspects of perception of juiciness and toughness was compared with Single-Attribute Time-Intensity (SATI) and Line Scale Profile.

Results showed that DATI provided a good separation of attributes and was equal to or better than SATI in differentiating beef samples based on perceived juiciness and toughness. By reducing the dumping effect and the inherent sample to sample variability, this method enabled more precise assessment of the relationship between juiciness and toughness in meat than SATI.

Download - 536 KB PDF
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Meat Science; Temporal Methods; Texture Analysis;

Duizer, L.M., Bloom, K., Findlay, C.J. Food Quality and Preference, 1997, 8(4), 261-269. Refereed Publication.

Dual-attribute time-intensity was evaluated as a method for the collection of the perception of two attributes simultaneously. Perceptions of sweetness and peppermint flavour within chewing gum were measured by 10 trained time-intensity panellists using both single-attribute and dual-attribute time-intensity evaluation.

In general, dual-attribute time-intensity was as sensitive as single-attribute testing in distinguishing between the sweetness and peppermint perceptions of chewing gum. In comparison to the single-attribute test, the dual-attribute test required half the time to complete and provided a means of assessing complex taste interaction during mastication. The dual-attribute test can be used to study relationships between two attributes within food products which possess a large degree of sample variability, such as the tenderness and juiciness of meat.

Download - 288 KB PDF
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Temporal Methods; Texture Analysis;

Swatland, H.J., Findlay, C.J. Food Quality and Preference, 1997, 8(3), 233-239. Refereed Publication.

The main muscles of commercially competitive cuts of beef (n=16) from a variety of sources were probed to detect ultraviolet (UV) fluorescence of connective tissue, together with a dynamic analysis of electromechanical signals for overall toughness. The main muscles were cut into 1.2cm cubes after being frozen. Muscle cubes were cooked for 20 minutes to an internal temperature of 70 ºC and evaluated by a trained panel.

Dynamic analysis showed that tough regions of meat cuts had a relatively high frequency of narrow fluorescence peaks subtending a small area under the fluorescence signal. Thus, for probe measurements made perpendicularly across muscles, the area under the fluorescence signal was correlated (p<0.01) positively with tenderness (r=0.57), and negatively with chewiness (r=-0.61) and residual tissue (r=-0.58).

Thus, despite variation in post-mortem treatment and cooking, connective tissue toughness may contribute to the overall toughness of commercially competitive cuts of beef, and sensory responses may be partly predicted from rapid, relatively non-destructive measurements on the raw meat.

Download - 6.4 MB PDF
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Descriptive Analysis; Meat Science; Sensory Instrumental Relationships; Texture Analysis;

Zimoch, J., Gullett, E.A. Food Quality and Preference, 1997, 8(3), 203-211. Refereed Publication.

This study evaluated temporal differences amongst panellists in perception of juiciness and tenderness of beef samples and explored the temporal relationship between juiciness and tenderness. Ten panellists evaluated samples from 48 animals using CSA computerized time-intensity (TI) procedure.

Grouping panellists for perception based on chewing behaviour using CSA curves was possible. Use of Principal Component Analysis (PCA) to produce curves based on PC scores over time provided more information about the samples and perception variability than simple averaging. Perception of tenderness was influenced by perceptual differences amongst panellists, and by the stage in mastication at which juiciness was perceived in a sample.

Download - 664 KB PDF
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Meat Science; Temporal Methods; Texture Analysis;

Duizer, L.M., Bloom, K. Findlay, C.J., Journal of Food Science, 1996, 61(3), 636-638. Refereed Publication.

The relationship between duration and maximum intensity of sweetness and peppermint flavour of chewing gum was explored using dual-attribute time-intensity sensory evaluation. Four chewing gum samples, varying in rate of release of sweetness and peppermint flavour were evaluated by 10 trained time-intensity panellists.

Chewing gum with a fast release of sweetness and peppermint flavour provided the highest maximum intensity and longest duration of sweetness and peppermint perception. The rate of release of sweetness was more important than rate of release of peppermint flavour in affecting duration of attributes.

Download - 360 KB PDF
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Temporal Methods;

Duizer, L.M., Gullett, E.A., Findlay, C.J. Meat Science, 1996, 42(2), 215-224. Refereed Publication.

The relationship between the perception of tenderness, chewing activity and instrumental compression was explored by time-intensity, electromyography and instrumental texture profile analysis (ITPA). Bovine m. longissmus dorsi from five treatments were evaluated by seven individuals.

Time-intensity results showed that the Decrease Area and Area Under the Curve provided the most information regarding sample differences, with the former providing the best sample discrimination. Electromyographic results of mastication rate demonstrated the number of chews required to reach maximum force to chew. The results suggest a need to re-examine the effects of early mastication vs. the late mastication effects for the measurement of meat tenderness.

Download - 644 KB PDF
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Meat Science; Sensory Instrumental Relationships; Temporal Methods; Texture Analysis;

Bloom, K., Duizer, L.M., Findlay, C.J. Journal of Sensory Studies, 1995, 10(3), 285-294. Refereed Publication.

A measure of the reliability (T-IR) of time-intensity measurements was developed based on the concept of standard deviation as a measure of panellist variability. The T-IR measure was applied to time-intensity data collected from 10 panellists evaluating the sweetness of 4 model sweetener solutions on horizontal and vertical time-intensity line orientations.

T-IR scores showed that the panellists were similarly reliable across the sweeteners and orientations. As well, independent of scale orientation, responses to sweeteners were similarly reliable. The T-IR measure can be used to maintain a high level of performance by monitoring time-intensity panellists. T-IR also provides an objective method of selecting panellists for time-intensity panels.

Download - 520 KB PDF
Categories: Analytical Sensory Methods; Statistical Methods and Data Visualization; Temporal Methods;

Duizer, L.M., Bloom, K., Findlay, C.J. Food Quality and Preference, 1995, 6(2), 121-126. Refereed Publication.

A trained time-intensity panel was used to evaluate the effect of scale orientation on time-intensity responses. Equi-sweet samples of aspartame, acesulfame k, sucralose and 9% sucralose were presented to 10 panellists for evaluation on both horizontal and vertical scales. For the most part, horizontal and vertical scales yielded similar results. However, Maximum Intensity responses on the vertical scale were approximately 13% greater than Maximum Intensity responses on the horizontal scale.

The parameters of Decrease Angle, Decrease Area and Area Under the Curve were also significantly larger when vertical scales were used than when horizontal scales were used. We suggest that differences can be minimized by anchoring reference samples to the scales and by counterbalancing the presentation of the scales within and amongst panellists. These results demonstrate the use of time-intensity scales on two dimensions and suggest the possibility of multi-attribute evaluations of taste.

Download - 468 KB PDF
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Temporal Methods;

Duizer, L.M., Gullett, E.A., Findlay, C.J. Journal of Food Science. 1993, 58(5), 943-947. Refereed Publication.

The Time-Intensity technique for measuring tenderness of bovine psoas major, longissimus dorsi, semitendinosus and shank was assessed. From the Time-Intensity curve, the Duration and area parameters (Increase and Decrease Area and Area Under the Curve) were most useful for sample separation. Using various Time-Intensity curve parameters, panellists were classified according to their perception of tenderness, with two clusters identified. A comparison of line scale results of force to chew and time to chew to the Time-Intensity resulted showed that comparable tenderness measurements were obtained by the two tests.

Download - 484 KB PDF
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Meat Science; Temporal Methods; Texture Analysis;

Barbut, S., Findlay, C.J. Critical Reviews in Poultry Biology, 1989, 2(1), 59-95. Refereed Publication.

Note: This publication is currently unavailable for download.
Categories: Meat Science;

Findlay, C.J., Stanley, D.W., Gullett, E.A. Meat Science, 1986, 16(1), 57-70. Refereed Publication.

Differential scanning calorimetry (DSC) was used to follow the three major endothermic transitions. (T1, T2, and T3) of beef muscle during heating. Borchardt and Daniels reaction kinetics were used to predict the three time and temperature treatments required to sequentially eliminate each transition. Longissimus dorsi and semimembraneous muscles were removed from beef carcasses suspended by Achilles tendon or pelvis.

Samples prepared by heating for 5 min at 57 ºC (I), 70 ºC(II) and 81 ºC(III) were assessed by sensory panel for tenderness, juiciness and residual connective tissue. Weight loss, Warner-Bratzler (W-B) shear and microstructure using transmission electron microscopy (TEM) were also determined. The I treatment showed a significant difference in tenderness and residual connective tissue between muscles, but not between contraction states.

The II treatment produced collagen shrinkage and a significant drop in W-B shear and residual connective tissue, coupled with increased tenderness in semimembraneous muscle. An increased W-B value, decreased juiciness, increased weight loss and a reduction in sarcomere and A-band length accompanied the III transition.

Muscles from carcasses that had been suspended by the pelvis were found to be significantly more tender than the same muscles from Achilles hung carcasses. It is concluded that DSC is capable of determining amount of protein denaturation and, hence, degree of cooking.

Download - 820 KB PDF
Categories: Analytical Sensory Methods; Differential Scanning Calorimetry; Meat Science; Sensory Instrumental Relationships; Texture Analysis;

Findlay, C.J., Gullett, E.A., Genner, D. Journal of Sensory Studies, 1986, 1(3-4), 307-314. Refereed Publication.

A computerized sensory analysis system, based on an IBM-PC compatible local area network, was developed. Panellist input was simplified through the use of a light pen and interactive questionnaire program. The system was integrated to allow preparation of descriptive, hedonic, triangle, structured and unstructured ballots; registration of panellists; collection of data; statistical analysis and report generation. The primary benefits are the simplicity of response for panellists, flexibility for the sensory analyst to design questionnaires and the elimination of time-consuming manual scoring and data manipulation involved in conventional sensory analysis.

Download - 420 KB PDF
Categories: Analytical Sensory Methods; Computerized Sensory Analysis;

Findlay, C.J., Stanley, D.W. Journal of Texture Studies, 1984, 15(1), 75-85. Refereed Publication.

Warner-Bratzler shear, Instron compression and extension were screened for sensitivity to the textural changes caused by heating scallop (Placopecten magellanicus) adductor muscle. Instron compression, expressed as hardness, was selected since it gave the greatest slope with respect to temperature. Previously frozen commercial scallop were heated to internal temperatures of 25 to 80 ºC. A linear increase in hardness of 0.033 N/g/ºC between 60 and 65 ºC which is considered to be a result of denaturation of myofibrillar proteins.

Hardness continued to increase above 65 ºC at a rate of 0.055 N/g/ ºC. Quantitative scanning electron microscopic (SEM) measurement of the proportion of irregular muscle fibers, expressed as a % damage, was performed on the same heated scallop used for texture analysis. Scallop heated from 25 to 50 ºC exhibited 30% damaged fibers; from 55 to 65 ºC, damage increased from 45% to 63%, paralleling the increase in hardness. Above 65 ºC, damage reached a maximum of 70%. The relationship between hardness and damage fit a linear model, with an R2 of 0.86 (P=0.004); thus, the microstructural measurement of damage to scallop muscle can be used to predict the textural property of hardness.

Download - 516 KB PDF
Categories: Texture Analysis;

Findlay, C.J., Stanley, D.W. Journal of Food Science, 1984, 49(6), 1513-1516. Refereed Publication.

Differential scanning calorimetry (DSC) was used to follow the changes in the endothermic transitions of beef muscle during conditioning. Sternomandibularis muscle held at 5ºC from 2-8 days post-mortem resulted in a significant (P < 0.05) drop in total heat of transition (ΔH) from 3.8 to 3.0 J/g. The myosin transition decreased from 57.8º to 55.2ºC while the actin transition increased from 81.8º to 83.2º (P < 0.05). Storage time and temperature were varied to generate a response surface of thermal data for psoas major and semimembraneous muscle. The decrease in ΔH of psoas major was optimal between 10 º and 13 ºC. Total ΔH of semimembraneous (3.9 J/g) was significantly greater (P < 0.05) than that of psoas major (3.4 J/g).

Download - 456 KB PDF
Categories: Differential Scanning Calorimetry; Meat Science;

Findlay, C.J., Stanley, D.W. Journal of Food Science, 1984, 49(6), 1529-1531, 1534. Refereed Publication.

Contraction state of beef muscle at onset of rigor influences tenderness of cooked meat. Loss in tenderness during cooking has been related, through use of differential scanning calorimetry (DSC), to thermal denaturation of myofibrallar proteins. Contraction of beef sternomandibularis muscle was controlled at sarcomere lengths of 2.4, 2.1, 1.9, 1.7, and 1.4 μm. Samples were scanned from 25-105 ºC at 10 ºC/min; ΔH (change in heat of transition) between 45 º and 92 ºC dropped from ca. 4 J/g muscle at 2.4 μm to ca. 3 J/g at 1.4 μm. This difference (P < 0.05) amounts to less than 1% of the total energy required to heat meat from 45 º to 92 ºC. The decrease is attributed to a greater actomysin contribution to the overall thermal curve resulting from increased overlap of the filaments.

Download - 380 KB PDF
Categories: Differential Scanning Calorimetry; Meat Science;

Meullenet, J.-F., Xiong, R., Findlay, C.J. Blackwell Publishing: 2007. Book.

From the Publisher:
Sensory scientists are often faced with making business decisions based on the results of complex sensory tests involving a multitude of variables. Multivariate and Probabilistic Analyses of Sensory Science Problems explains the multivariate and probabilistic methods available to sensory scientists involved in product development or maintenance. The techniques discussed address sensory problems such as panel performance, product profiling, and exploration of consumer data, including segmentation and identifying drivers of liking.

Applied in approach and written for non-statisticians, the text is aimed at sensory scientists who deal mostly with descriptive analysis and consumer studies. Multivariate and Probabilistic Analyses of Sensory Science Problems offers simple, easy-to-understand explanations of difficult statistical concepts and provides an extensive list of case studies with step-by-step instructions for performing analyses and interpreting the results.

Coverage includes a refresher on basic multivariate statistical concepts; use of common data sets throughout the text; summary tables presenting the pros and cons of specific methods and the conclusions that may be drawn from using various methods; and sample program codes to perform the analyses and sample outputs.

As the latest member of the IFT Press series, Multivariate and Probabilistic Analyses of Sensory Science Problems will be welcomed by sensory scientists in the food industry and other industries using similar testing methodologies, as well as by faculty teaching advanced sensory courses, and professionals conducting and participating in workshops addressing multivariate analysis of sensory and consumer data.

Categories: Analytical Sensory Methods; Consumer Research; Descriptive Analysis; Statistical Methods;

Findlay, C.J., Wilson, H., Spears, M., Cowen, S., Castura, J.C. 8th Pangborn Sensory Science Symposium. July 26-30, Florence, Italy. Scientific Presentation (Poster).

In a random sampling of consumers, it is not unusual to have a proportion of the panellists who are neither users or purchasers of the product. This means that “liking” responses to products are not informed by either context or experience. This reduces the validity of the test. When we consider that choice reflects the ability to detect a difference, in general population consumer difference tests the accepted proportion of differentiators (Pd) is only 30%. To increase the validity of consumer tests, particularly in the case of product matching, it is important to gather the responses of actual product consumers and if possible, heavy users.

In standard recruitment, consumers are asked to identify their own purchase and usage behaviour. It is commonly recognized that consumers will answer these questions based upon their recollection or response to the desire to participate in the test. In short, consumers lie; this compromises the quality of the data collected.

By recruiting consumers on the basis of purchase history, it is possible to increase the proportion of consumers in the panel who are real customers. The information can be gained from actual purchase history. The usefulness of this information was tested in two ways. Previous studies were examined to determine which of the consumers who were part of the test were verified as regular purchasers. Non-users and regular purchasers were grouped and compared on the basis of their historic data. The range of products evaluated by about 100 consumers each were; Bath Essence, Deodorant, Laundry Tablets and Food Wrap. Similar conclusions and power of the test were obtained with 20 to 35% fewer regular customers. A verification study with Tesco Home Panels using two groups of consumers was performed on a selected non-food product (Laundry Tablets). One group was composed of proven product purchasers. The second group was made up of proven non-purchasers. Their liking results were analyzed for both mean results and variance. Power tests performed on random subsets of the data demonstrated that smaller panels of well selected consumers reliably deliver the same outcome.

Note: This publication is currently unavailable for download.
Categories: Consumer Research; Web-based Consumer Testing;

Castura, J. C. 2nd Meeting of The Society of Sensory Professionals. October 27-29, 2010. Napa, CA, USA. Scientific Presentation (Poster). (Forthcoming).

A choose-all-that-apply (CATA) question allows respondents to select multiple answers from a list. A technique called answer piping displays the respondent’s selections as possible responses in a subsequent question. Answer piping was used to allow consumers to shortlist wine-producing regions before ranking those regions for quality in a Chilean red wine consumer study conducted in fall 2007. Red wine consumers were recruited on 11 occasions in the aisles of five LCBO stores in Toronto and nearby cities. 614 consumers whose in-store shopping activity revealed particular purchase intentions were invited into tasting rooms to evaluate 3 red wines. During delays between samples, consumers were asked demographic, attitude, and usage questions as part of the web-based questionnaire presented on tablet computers. The Compusense at-hand questionnaire used answer piping on several occasions. In one CATA question, consumers selected the wine-producing regions which they associated with high quality wines. Consumers then ranked the wine regions that they had selected as producing high-quality wines. Ties were allowed. Rank sums were calculated from the recoded and merged ordinal data. Consumers reduced the average number of wine regions to be ranked from 14 to 4.93 (s=2.91). The median number of wine regions ranked was 4. The mode was 2 (144 consumers); 4 consumers selected all 14 regions. Results indicate that overall the consumers perceived Australia highest for quality, followed by France, Italy, Canada, California, and Chile. California was ranked higher for quality by consumers who had shortlisted 5 or more wine regions than by those consumers who shortlisted fewer wine regions. These 6 wine-producing regions were listed as consumers’ top regions for purchase. LCBO sales data indicates that top wine regions by net sales and volume for 2007-2008 were Canada, Italy, Australia, France, U.S. and Chile. Canadian wines were sold at the lowest price per volume.

The 15th IUFoST Congress, Parallel Session: Sensory Science, Physical Science and Product development, August 22-26, 2010. Cape Town, South Africa. Scientific Presentation (Oral). (Forthcoming).

The Feedback Calibration Method (Compusense® FCM) has been proven to cut training time of descriptive panels in half while attaining optimum proficiency. Panels trained using FCM provide stable and consistent analytical descriptive analysis results. This method ensures that sensory profile data collected over time (shelf-life or category assessments) and across different locations (regional panels or global organizations) is calibrated and completely comparable.

The FCM method is well-grounded in the category-learning research of Dr. Greg Ashby, UC Santa Barbara, whose studies prove that immediate feedback (within 2.5 seconds) produces a dopamine release that physically reinforces the synapses in the brain. This type of reinforcement of category learning has been shown to deliver long-term benefits that allow very rapid refreshing of prior training. The net effect is that it is easier to train panellists and it is much easier to retrain them in a product area after a break from active panel work.

The training method uses the sensory order of operations approach to the development of descriptive panels. With all attributes clearly defined by the panel in their training, there is no ambiguity in panellist performance.

Feedback training develops the individual panellist skills in use of the ballot and the scale. The immediate reinforcement of descriptive panellist learning delivers greater proficiency in a much shorter period of time. Typically, a new panel that has undergone basic screening can be trained to proficiency in a product area within 20 hours over 10 training sessions. Retraining takes as little as one 2 hour session. This approach to descriptive analysis is compatible with elements of the Spectrum® method and of QDA®. The application of this approach to the training of wine descriptive panels will be presented.

Castura, J. C. 10th Meeting of the Sensometric Society. July 25-28, 2010. Rotterdam, The Netherlands. Scientific Presentation (Oral). (Forthcoming).

Sensory professionals seeking guidance in best practices often turn to publications from standards organizations such as ISO and ASTM. A review of guides related to sensory equivalence testing will be presented. In several cases the power approach is prescribed for determining equivalency, but this approach is problematic. It attempts to control beta risk in the difference test and declare samples equivalent when the null hypothesis is retained. Its ineffectiveness in practice can be demonstrated through simulations following the approach used by Bi (2005). Code was implemented independently in R and simulations obtained. For example, triangle and duo-trio test results were simulated where the true proportion of discriminators was set to 0.1. The power approach confirmed similarity in the triangle test with probability 0.6005 when the sample size was 54 and with probability 0.0278 when the sample size was 540. The power approach confirmed similarity in the duo-trio test with probability 0.6273 when the sample size was 96 and with probability 0.0323 when the sample size was 960. As precision of measurement increased (with increasing sample size) the probability of concluding that samples were equivalent was diminished, which further underscores the failure of the power approach. Until standards and guidelines can be updated, practitioners must look elsewhere for direction when conducting equivalence tests. Results are contrasted with other statistical approaches. For special cases investigated the sensR package (Christensen & Brockhoff, 2010) implements methods related to equivalence testing where power calculations correspond more closely to simulated results.

Castura, J.C. ASTM E-18.04 Seminar on Discrimination Methods. April 22, 2010. St. Louis, MO, USA. Scientific Presentation (Oral). (Forthcoming).

This abstract is unavailable at this time. Please contact us for additional information.

Findlay, C.J., Meullenet, J.-F., and McNicholas, P. 8th Pangborn Sensory Science Symposium. July 26-30, 2008,Florence, Italy. Scientific Presentation (Poster).

Consumer testing of products which create sensory fatigue have a number of serious challenges. The effect of consumption of alcoholic beverages, extremely spicy foods, intense flavors or numbing ingredients limit the collection of complete block data to a small number of samples. If a study with a large number of samples is conducted by collecting consumer data over several days, learning affects the quality of the consumer responses. By the third day, most consumers are behaving like trained assessors, a conclusion that is supported by the decrease in first position effect.

Typically, segmentation of consumer liking data requires a complete block. In this study, 12 commercially available Cabernet Sauvignon wines were evaluated by over 600 red wine consumers in a 12 present 3 Balanced Incomplete Block design. Each consumer tasted 3 of the wines in a single 10 minute session, with demographic questions providing a break between samples. The means for all 12 wines ranged from 5.7 to 6.3 on the 9-point hedonic scale. Without segmentation of the consumers the results were not actionable. It was essential to determine clusters based on consumer liking. But a complete block of data had to be created. To compensate for the missing data points, the average response of each panellist was inserted into the nine missing data points. The total data set was treated by Qannari Clustering (Senstools 3.3.1).

Two additional mathematical approaches were used to cluster the data and provided comparable conclusions. Descriptive sensory data of the clustered products provided an external validation of the selection of four consumer liking segments. The clusters ranged in size from 17 to 32% and each had distinct sensory differences that were understood by winemakers. The liking range within each cluster expanded to around 4 to 8 on the 9-point scale. This approach to segmentation of large BIB studies with small incomplete blocks that combine sensory driven design with a specific clustering procedure appears to be very promising.

Castura, J.C. 8th Pangborn Sensory Science Symposium. July 26-30, Florence, Italy. Scientific Presentation (Oral).

A so-called donkey voter selects candidates according to position on an election ballot. Are untrained sensory panellists similarly influenced by position when responding to choose-all-that-apply (CATA) questions? In sensory and consumer testing, lists of choices, conventionally presented in fixed order, allow panellists to indicate sensory perceptions without requirements for scaling. Results help in understanding products and drivers of hedonic response.

Using Compusense at-hand, colleagues at University of Arkansas and Compusense Inc. presented 10 commercial orange juices to 106 student panellists. Separate CATA questions were presented for different sensory modalities as follows: 5 appearance choices (one column), 28 flavour choices (3 columns of 10, 10, and 8), and 9 texture choices (1 column). In each case “none of these apply” appeared in the final position (quite different and rarely selected, it was omitted from analysis), and during sessions 3 and 4 all other choices were presented using a Williams design with choice sets assigned to sample sets. The Next button appeared at bottom right.

First positions increased selection percentage points for appearance (+5.9%), flavour (+2.6%) and texture (+2.8%). Attributes in the leftmost flavour column were chosen more than either those in middle (+2.5%) or rightmost (+3.6%) columns. First position added 10-20% selections by proportion. Various data adjustments were considered to confirm the absence of artifacts. Computerized visualizations were developed to vividly demonstrate results. Results raise strong concerns that fixed choice order ballots skew CATA results, with implications for anyone conducting sensory and consumer tests in this manner. Rotation of samples is commonplace in designed experiments, and rotation of choices, as performed in this study, is recommended for improving data quality.

Note: This publication is forthcoming.
Categories: Consumer Research; Web-based Consumer Testing;

Meullenet, J.-F., Findlay, C.J., Tubbs, J.K., Laird, M., Kuttappan, V.A., Tokar, T., Over, K., Lee, Y.S. 8th Pangborn Sensory Science Symposium. July 26-30, Florence, Italy. Scientific Presentation (Poster).

Check-all-that-apply (CATA) questions have been used in consumer studies to determine key sensory attributes characterizing a specific product. CATA has the particularity of assessing perceived product attributes without requiring scaling. The objective was to determine the effects of the number and order of the choices in CATA questions on attribute selection and consumer response time.

Ten commercial orange juices (OJ) were presented to 106 consumers. The tests were conducted over two weeks in two sessions for each week. Consumers were given CATA questions to describe appearance, flavor and texture each with 5, 27, and 11 descriptors, respectively. This allowed the investigation of response time as a function of number of choices. Effects of choices presentation order (alphabetical for week1 and Williams design for week2) on the OJ sensory descriptions were also examined. The study was designed, organized and administered using Compusense® at-hand (Compusense Inc., Guelph, Canada).

Consumer response time revealed that for the William design presentation order of choices, consumers took in average 4.54, 5.00 and 1.69 seconds more to answer CATA questions pertaining to appearance, flavor and texture attributes, respectively. However, product descriptions showed no significant differences between the designed and alphabetical presentations. Consumer response time was also investigated as a function of sample presentation order (Fig. 1). Not surprisingly, response time for CATA questions decreased as a function of sample presentation position within a session, showing the same effect on day2 of each week. Presentation position also had an impact on the number of choices made. The average number of descriptors chosen for flavor increased from 4.2 for the first sample tested to 4.8 for the 5th sample tested during week1/session1.

In conclusion, the time taken by consumers to answer CATA questions is impacted by individual variations and the order of the response options are presented in. However, overall product descriptions were not impacted by CATA descriptors presentation order.

Note: This publication is forthcoming.
Categories: Consumer Research; Web-based Consumer Testing;

Findlay, C.J. 9th Sensometrics Meeting. July 20-23, 2008, St. Catharines, Canada. Scientific Presentation (Oral).

Consumer testing of beverage alcohol has a number of serious challenges. The effect of consumption of alcohol is a limiting factor in obtaining complete block data. Collecting consumer data over several days affects the quality of the consumer response. By the third day, most consumers are behaving like trained assessors, a conclusion that is supported by the decrease in first position effect. Typically, segmentation of consumer liking data requires a complete block.

In this study, 12 Cabernet Sauvignon wines were evaluated by over 600 red wine consumers in a 12 present 3 Balanced Incomplete Block design. Each consumer tasted 3 of the wines in a single 10 minute session, with demographic questions providing a break between samples. A total of 11 sessions were conducted at 5 LCBO store locations.

Three approaches were used to provide dummy variables for the missing data in each set. The average response for the panellist was inserted into the missing data points. The product average was substituted in a second data analysis and finally the overall mean was used in the third data set. Each approach was subjected to Qannari Clustering (Senstools 3.3.1) and 3, 4 and 5 Cluster solutions were considered. Grouping of products based on descriptive sensory data provided an external validation of the selection of sensory segments. A four cluster solution using the panellist mean produced clusters that were well explained by the sensory contrasts.

Note: This publication is currently unavailable for download.
Categories: Consumer Research; Statistical Methods and Data Visualization; Web-based Consumer Testing; Wine;

Findlay, C.J. The Institute of Food Technologists Annual Meeting + Food Expo 2008. Technical program: Sensory Evaluation: Sensory assessments outside the mouth. July 28-July 1, 2008. New Orleans, LA, USA. Scientific Presentation (Oral).

This abstract is unavailable at this time. Please contact us for additional information.

Note: This publication is currently unavailable for download.
Categories: Analytical Sensory Methods; Texture Analysis;

Findlay, C.J., Phipps, K. S., Pitts, S., Fortune, S., Moore, L., Castura, J.C. 7th Pangborn Sensory Science Symposium. August 12-16, 2007, Minneapolis, MN, USA. Scientific Presentation (Poster).

The cost and time required for training descriptive analysis panels is often cited as a major barrier to the routine application of descriptive sensory analysis. Compusense FCM® was developed as a method to accelerate the training of descriptive panels and to provide
a mechanism for calibration that would stabilize descriptive analysis data over time and across panels.

By providing individuals with immediate and accurate feedback during training sessions, these panels have been shown to require fewer sessions to achieve accuracy and precision comparable to well-trained conventional panels (Findlay et al., 2006; Findlay et al., 2007). Previously published research on the feedback calibration method (Compusense FCM) was designed to test specific hypotheses on the performance of panels as a whole. This case study addresses the effect of the routine application of Compusense FCM on panellists who are members of ongoing descriptive panels.

Download - 1.9 MB PDF
Categories: Analytical Sensory Methods; Compusense FCM; Computerized Sensory Analysis; Descriptive Analysis;

Castura, J.C., Findlay, C.J. 7th Pangborn Sensory Science Symposium. August 12-16, 2007, Minneapolis, MN, USA. Scientific Presentation (Poster).

Training sessions often yield a limited dataset, which in turn restricts available analyses. Gathering ideal data sets for analysis might be at odds with imperatives of training regimen. Raw data is too voluminous to consider in numerical form. Humans have excellent ability for pattern recognition. Multifunctional graphs can reveal both macro and micro structures in the data.

Download - 122 KB PDF
Categories: Analytical Sensory Methods; Compusense FCM®; Computerized Sensory Analysis; Descriptive Analysis; Statistical Methods and Data Visualization;

Castura, J.C., Findlay, C.J. A Sense of Diversity: European Conference on Sensory Consumer Science of Food and Beverages. September 26-29, 2006, The Hague, The Netherlands. Scientific Presentation (Poster).

Descriptive analysis is applied to a diverse range of complex, real-world food and consumer products because the information it provides about those products is unrivalled in its richness. A common lexicon allows the descriptive sensory panel to reference sensory attributes of products undergoing study in a highly specific and consistent manner. When combined with best practices it eliminates ambiguity of response and yields highly actionable results. A range of sensory attributes relevant to and representative of the product category undergoing study is typically selected.

Sensory attributes are often categorized using non-sensory frameworks: chemistry (e.g. limonene might be described first by its chemical properties, noting its lemon-orange or turpentine-pine odor, depending on chirality) or physical or biological origin (e.g. "notes" from banana and apple to lime and grapefruit, might be categorized as "fruity").

An opportunity exists for systematic classification of sensory attributes based on the difficulty associated with (i) identifying and (ii) scaling the attribute in specific product contexts. Diverse sensory attributes could be grouped into four broad categories (easy to identify and scale, easy to identify but hard to scale, hard to identify but easy to scale, and hard to identify and scale). The authors are unaware of any attribute classification system with this kind of sensory approach.

Data from a red wine descriptive sensory panel (130 attributes) and two white wine descriptive sensory panels (110 and 76 attributes were used by a previously trained and previously untrained panel, respectively) allowed the authors to classify attributes. Resultant classifications will provide subsequent panel leaders with insights into the types of responses that might be expected from future descriptive sensory panels, influence subsequent ballots, and permit for testing of predictions related to existing attribute classifications.

Download - 116 KB PDF
Categories: Analytical Sensory Methods; Descriptive Analysis;

Castura, J.C., Findlay, C.J. 8th Sensometrics Meeting. August 2-4, 2006, Ås, Norway. Scientific Presentation (Oral).

Descriptive analysis provides valuable information about the sensory properties of consumer products, but this information lacks the temporal dimensionality of real-world sensory experiences. Type II error occurs when the descriptive sensory panel fails to differentiate between products known to be discriminable. Findlay (2000) reported no meaningful reduction in beta risk when descriptive analysis on manipulated salad dressings was augmented by order of initial perception data. Attributes on the ballot appeared in fixed order but panellists were required to respond in order of perception, and responding to order of perception increased the complexity of descriptive analysis. Pineau (2004) discussed the Temporal Dominance of Sensations (TDS) methodology, which also recognizes the promise of temporal data.

Guessing models for difference tests assume that correct responses are given by discriminators, who perceive true differences, and non-discriminators, who guess correctly. Panellists might adopt strategies that enhance discriminative ability in a triangle test (Rousseau, 2001), although experimenters cannot verify strategy use.

Enriched datasets might provide insights in these and other areas. Computers facilitate data collection, and have potential to gather temporal metadata, providing contextually enriched data without detriment to existing analyses. Enriched datasets might include irregularly spaced temporal data representing discrete, sometimes dependent events that might better model and provide insights into sensory experiences. Myriad opportunities for investigation arise.

A descriptive panel might respond to attribute intensity multiple times – leaving their task largely unchanged – permitting supplemental analysis of either incomplete or interpolated data within specific time intervals. Relationships between response time, decision, and accuracy in difference testing might provide new information. Relationships between correctness and decisiveness in descriptive panel training also merit exploration. Consumers might be grouped for analysis according to temporal response patterns.

To assess the potential value of temporal metadata, twelve panellists were selected and instructed to respond to dual attribute time intensity (DATI) where anchors were "Same" and "Different". Samples A and B were Premium salted soda crackers (32.6 mg Na / cracker) and Premium unsalted soda crackers (21.4 mg Na / cracker), respectively. Each panellist received four pairs (AA, AB, BA, BB) according to Williams Latin square design (four treatments).

There were 40/48 (83%) correct identifications (p<0.001). Panellist indicated "same" after an average of 23.2 s (sd=13.3 s) when correct and 25.3 s (sd=16.3 s) when incorrect. Panellists indicated "different" after an average of 14.5 s (sd=9.3 s) when correct and 17.7 s (sd=10.4 s) when incorrect. Panellists took significantly longer to declare "same" than "different" according to one-way ANOVA (p=0.019) in this preliminary test: further investigations are planned to determine relationships between d', decision, correctness, and response time.

Computerized sensory and consumer data collection systems allow rapid implementation of tests based on standard methodologies, cost reductions in the testing process, global reach, and increased throughput. Historical data, rich with temporal and other contextual information, could further increase utility by providing a relationship-rich data store, which can be mined in interesting ways, and provide a basis for knowledge discovery.

Note: This publication is currently unavailable for download.
Categories: Analytical Sensory Methods; Difference Testing & Equivalence Testing; Temporal Methods;

Findlay, C.J., Phipps, K., Castura, J.C., 6th Pangborn Sensory Science Symposium. August 7-11, 2005. Harrogate, Yorkshire, UK. Scientific Presentation (Poster).

Compusense FCM® (feedback calibration method) has been shown to be an effective tool to train descriptive analysis panels. The key to making this method work is providing “true” information, feedback, to panellists at the time they evaluate the attribute. This permits immediate calibration of the response. If the feedback is either trivial or incorrect, the panellist may be confused and the desired learning will not take place. To establish meaningful targets for feedback it is important to understand the shape of the psychometric function and the portion of the curve that describes the attribute intensity for the product being studied.

A specific attribute may be identified and defined in a range of the different products. The perception of the attribute will be dependent upon the product being tested. In simplest terms, the perceived sweetness of the same concentration of sucrose will be quite different in a citrus drink than in water. Both the just noticeable difference (JND) and the threshold values will be influenced by the components that make up any system. Sensory attributes can be assigned to several categories that can assist in applying the most appropriate strategy for both training and the collection of data.

The results from three large studies on wine will be used to illustrate the factors that influence the selection of both targets and ranges. The 76 to 130 attributes found in both red and white wines will be used to explain the feedback strategy for setting meaningful targets. By understanding the sensory dynamics of attributes, it is possible for sensory analysts and panel leaders to refine the process of training optimal panels.

Note: This publication is currently unavailable for download.
Categories: Analytical Sensory Methods; Compusense FCM®; Computerized Sensory Analysis; Descriptive Analysis;

Castura, J.C., Findlay, C.J., 6th Pangborn Sensory Science Symposium. August 7-11, 2005. Harrogate, Yorkshire, UK. Scientific Presentation (Poster).

Two panels, one composed of experienced red wine panellists (Panel T), the other of panellists without experience in sensory analysis (Panel U), were recruited and trained to evaluate white wine. Each used the Wine Aroma Wheel to develop white wine lexicons over five 2.5h training sessions. Panels T and U used 110 and 76 line scale attributes, respectively. Each panel established their own training targets based on 90% confidence intervals. Panels were calibrated with Compusense FCM®. Training targets were iteratively refined over 4 sessions. When training concluded each panel evaluated the same 20 white wines in triplicate. Permutation tests of the RV coefficient demonstrated strong similarity between the panels' product configurations in sensory space.

There were 46 attributes (29 aroma, 6 taste/mouth feel, 11 flavour) with similar or identical descriptors and reference standards. For each common attribute, panellist mean scores across all products were calculated. Centroid cluster analysis formed panellist groups consistent with panel membership, reflecting the panel-specific manner in which line scales were used. For each panellist, the scale distance between the maximum and minimum wine mean scores was then obtained for each attribute, and Fisher's LSD (p=0.05) calculated. Dividing range by LSD value reflected a panellist's ability to discriminate wines using the attribute; higher scores indicated greater differences being detected. Groups formed when these quotients were submitted to centroid cluster analysis did not reflect panel membership. Quotients calculated on panel mean scores showed Panel T had higher quotients for 20 of 46 attributes, further supporting the similarity in ability to detect differences.

Regardless of previous descriptive sensory training and calibration that resulted from panel-independent generation and refinement of training targets using Compusense FCM®, individual panellists performed similarly in detecting differences among wines. Furthermore, both panels produced meaningful product profiles and displayed similar abilities to detect differences.

Download - 292 KB PDF
Categories: Analytical Sensory Methods; Compusense FCM®; Computerized Sensory Analysis; Descriptive Analysis; Wine;

Findlay, C.J., The Institute of Food Technologists Annual Meeting + Food Expo 2005. Technical program: Sensory Evaluation: Analytical. July 17-20, 2005. New Orleans, LA, USA. Scientific Presentation (Oral).

Data from a descriptive analysis panel sometimes fails to detect differences between products for one or more sensory attributes. Results might nonetheless be consistent with the best possible data; lack of discrimination could be meaningful information if no true sensory difference exists between products for the attribute, which might occur when all products fall within one just-noticeable difference interval (Castura et al., 2006).

Download - 328 KB PDF
Categories: Analytical Sensory Methods; Compusense FCM^®; Computerized Sensory Analysis; Descriptive Analysis; Wine

Findlay, C.J., A Sense of Identity: European Conference on Sensory Consumer Science of Food and Beverages. September 26-29, 2004. Florence, Italy. Scientific Presentation (Poster).

The region of the Niagara Peninsula of Ontario Canada, has developed into a significant producer of varietal wines. New world winemakers are caught between the desire to produce imitations of the original examples of the varietal products and wines that express their distinctive terroir. The popularity of white wines, particularly Chardonnay, has led to its production in most of the region’s 80 wineries. This research examines the sensory properties of a selection of these wines compared to international products.

Download - 204 KB PDF
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Descriptive Analysis; Wine;

Lesschaeve, I., Findlay, C.J., 12th Australian Wine Industry Technical Conference proceedings. July 25-29, 2004. Melbourne, Australia. Scientific Presentation (Oral).

In an era of global market competition, wine companies realize the need to understand better consumer preferences and respond to their needs effectively. At the 11th Australian Wine Industry Technical Conference Terry Lee presented a paper (Lesschaeve et al. 2002) on the use of preference mapping to define successfully the sensory preferences of wine consumers. The current study proposes a strategy to target and develop a wine style based on preference mapping outcomes.

Twelve white wines were selected to represent a specific category available in Ontario liquor stores. One hundred and fifteen Canadian consumers from the Greater Toronto Area were recruited according to specific demographic criteria, as well as their white wine purchase and consumption habits. Consumers participated in tasting sessions held on three consecutive days.

During each session, they tasted four of the 12 selected wines according to a specific experimental design and indicated their overall liking. Eight of the twelve wines were then evaluated in triplicate by an extensively trained panel for a comprehensive range of sensory attributes. Sensory preferences were mapped using internal preference mapping techniques aimed at explaining the preference of consumers in terms of sensory attributes of the wine.

An opportunity for developing a new white wine style was highlighted. The profile of this new style was defined by its coordinates on the preference map. Then, the expected intensities of its sensory attributes were obtained by reverse engineering the coordinates into attribute scores (Moskowitz 1994). Strategies are proposed to communicate effectively the sensory profile of the new desired wine style to winemakers.

Download
Categories: Analytical Sensory Methods; Consumer Research; Descriptive Analysis; Wine;

Findlay, C.J., Castura, J.C., Schlich, P., Lesschaeve, I. 7th Sensometrics Meeting. July 28-30, 2004. Davis, CA, USA. Scientific Presentation (Oral).

The performance of descriptive panels is typically determined by post-hoc data analysis. Poor panel performance is determined after the fact and arrives too late to help the panel leader in training. The Feedback Calibration Method (FCM^®) is an effective method for training descriptive panellists. FCM optimizes proficiency by ensuring efficient panel training.

Two panels were recruited and trained to evaluate white wine; one panel was composed of experienced red wine panellists (Panel T), the other of panellists with no experience in sensory analysis (Panel U). Each panel used the Wine Aroma Wheel to develop their own white wine lexicon over 5 days of training sessions of 2.5h each. Panels T and U used 110 and 76 line scale attributes, respectively. Four additional training sessions were used to apply best practices from conventional training and computerized feedback. Training targets were based on 90% confidence intervals around the mean values on line scales anchored at 0 and 100. The panels refined their own training targets iteratively. At the conclusion of training, each panel evaluated the same 20 white wines in triplicate.

Permutation tests of the RV coefficient were used to compare the panels in terms of the underlying sensory space. The results of the panels were similar, and both Panel T and U were superior to a proficient conventionally trained red wine panel (Panel D). Panel U performed similarly to Panel T on proportion of attributes discriminated and disagreement using a two-way mixed-model analysis of variance and on multivariate discrimination evaluated by a MANOVA with the same mixed model. Evaluation means for product*attribute fell within the training range targets in 59% of the cases for Panel T and 68% for and U, providing an indication of the panels' abilities to hit the training targets. Panel U was shown to be proficient (p=0.05) after only 9 formal training sessions (22.5h), a reduction in training time of 48.75%.

Note: This publication is currently unavailable for download.
Categories: Analytical Sensory Methods; Compusense FCM®; Computerized Sensory Analysis; Descriptive Analysis; Computerized Sensory Analysis; Wine;

Castura, J.C., Findlay, C., Lesschaeve, I. 7th Sensometrics Meeting. July 28-30, 2004. Davis, CA, USA. Scientific Presentation (Poster).

Training targets can be established from product profiles that provide an objective representation of the underlying sensory characteristics of a group of products. If available to a panel leader, these training targets can be used to calibrate a panel and measure the accuracy of their responses.

Response accuracy can be determined either by frequency counts—how often the target was hit versus how many opportunities there were to hit it—or by distance from target measurements— which attempt to further quantify the degree to which the target was hit or missed. In addition to the frequency counts, four distance-from-target methods are presented and discussed—Distance from Target, Distance from Range, Adjusted Distance from Target, and Adjusted Distance from Range—each of which provides insights into the degree to which the panel is calibrated.

Download - 145 KB PDF
Categories: Analytical Sensory Methods; Compusense FCM^®; Descriptive Analysis; Statistical Methods and Data Visualization;

Findlay, C.J. The Institute of Food Technologists Annual Meeting + Food Expo 2004. Technical Program: Advances in sensory science symposium. July 12-16, 2004. Las Vegas, NV, USA. Scientific Presentation (Oral).

Descriptive sensory analysis is one of the most powerful tools available to the sensory scientist. Regardless of the individual approach to descriptive analysis there are the common steps of identifying the attributes that describe the product, bringing a panel to agreement on the descriptors that are used, establishing a working scale that captures the range of intensities and practicing the method to gain individual and collective proficiency. Much emphasis has been placed on the statistical measures of panel and panellist performance. Although this is important, it may only tell us after the fact that the panel was off target.

Optimization of the descriptive analysis panel focuses on training and providing immediate, meaningful feedback to accelerate learning and establish calibration standards permitting panel-to-panel comparisons over time and locations. Computerized feedback has been demonstrated to be an effective training tool. However, for the feedback to work it must also be true and consistent. This requires a clear understanding of the behaviour of the psychometric function of each attribute within the context of the product being tested. Optimization also relies on group feedback at the end of each session that reinforces the learning that takes place. The results of several panels will be used to provide specific examples of the power of this method.

Download - 631 KB PDF
Categories: Analytical Sensory Methods; Compusense FCM^®; Computerized Sensory Analysis; Descriptive Analysis; Wine;

Findlay, C.J., Bartel, A., Lesschaeve, I. 55th American Society for Enology and Viticulture Annual Conference. June 29-July 2, 2004. San Diego, CA, USA. Scientific Presentation (Poster).

This abstract is unavailable at this time. Please contact us for additional information.

Note - This publication is currently unavailable for download.
Categories: Analytical Sensory Methods; Compusense FCM^®; Computerized Sensory Analysis; Descriptive Analysis; Wine;

Findlay, C.J., Castura, J.C., Lesschaeve, I. 5th Pangborn Sensory Science Symposium. July 20-24, 2003. Boston, MA, USA. Scientific Presentation (Oral).

Descriptive analysis is one of the most powerful tools available to sensory scientists. However, regardless of the approach being used to analyze the sensory attributes of products, descriptive panels require significant training before the panel members, individually, and the panels collectively, become a reliable sensory instrument. There is great panel-to-panel variability and the training style of panel leaders can have a great influence on results. This research proposes the use of immediate feedback with calibration standards (feedback calibration) as a method to improve the training process and to provide anchors which permit comparison between panels.

An experienced determination panel performed descriptive profiling of 20 red wines. Their results were used to establish the attributes and targets for the second phase of the research. Sixteen inexperienced panellists were recruited and given 20 hours of common training over 10 days. They were then divided into two panels, control and experimental, composed of 5 women and 3 men each. The control panel was trained using conventional debriefing at the end of each session. The experimental panel only received immediate computerized feedback in the booths during evaluation. Both panels saw the same 10 wines and used the same scales and attributes. The research continued daily over a three-week period.

Extensive statistical analysis indicated that both the experimental and control panels were able to reproduce the results obtained by the determination panel. Panellist and panel accuracy and precision were obtained by measuring the difference from the target values. Both panels demonstrated similar learning curves. The conclusion from this preliminary work is that Feedback Calibration can provide unbiased and effective training for panellists, regardless of the style, skill or experience level of the trainer. Further research will be conducted to determine if the combination of both techniques would result in faster or more accurate descriptive panel training.

Download - 128 KB PDF
Categories: Analytical Sensory Methods; Compusense FCM^®; Computerized Sensory Analysis; Descriptive Analysis; Wine

Findlay, C.J., Castura, J.C., Lesschaeve, I. 5th Pangborn Sensory Science Symposium. July 20-24, 2003. Boston, MA, USA. Scientific Presentation (Poster).

Descriptive analysis is the most refined method for determining sensory information about the shelf life of products, the manufacturing processes used by competitors, and the attributes that drive consumer preference. The quality of the sensory profile depends largely on the extent to which the sensory panel – the analytical instrument of descriptive analysis – is trained. Calibrating descriptive sensory panels reduces unwanted variability. Feedback Calibration is an effective method for calibrating descriptive sensory panels. The Feedback Calibration method calibrates panellists by providing immediate feedback during the normal flow of a computerized ballot.

A sensory profile of 20 red wines produced by an experienced determination panel was used to establish attributes and targets for the second phase of the research. Sixteen inexperienced panellists were recruited and provided 20 hours of common training over 10 days. Panellists were then divided into two panels, control and experimental, composed of 5 women and 3 men each. The control panel was trained using conventional debriefing at the end of each session. The experimental panel received immediate computerized feedback in the booths during evaluation. Both panels saw the same 10 wines, and used the same scales and attributes. Ten additional training sessions were conducted over a three-week period. On the days following the 4th, 6th, 8th, and 10th training sessions, the panels evaluated 5 unfamiliar wines in the absence of feedback.

The experimental design allowed assessment of the impact on panel performance caused by shifting from a training environment (in which feedback was provided) to a testing environment (in which no feedback was provided). Results provided insights into the relative strengths of the FC method versus conventional training from the perspective of discrimination, distance-from-target, agreement among panellists, and panel agreement with expected attribute values.

Note: This publication is currently unavailable for download.
Categories: Analytical Sensory Methods; Compusense FCM^®; Computerized Sensory Analysis; Descriptive Analysis

Plaul, D., Findlay, C.J., Meier-Ploeger, A. 4th Pangborn Sensory Science Symposium 2001. July 22-26, 2001. Dijon, France. Scientific Presentation (Poster).

This abstract is unavailable at this time. Please contact us for additional information.

Note: This publication is currently unavailable for download.
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Descriptive Analysis; Meat Science;

Findlay, C.J. The Institute of Food Technologists Annual Meeting + Food Expo 2000. Technical program: Sensory Evaluation. June 10-14, 2000, Dallas, TX, USA. Scientific Presentation (Oral).

Analytical sensory profiles are used to guide new product development, match products, and determine the effect of raw material and process changes. Occasionally, products that are not significantly different in sensory profile are found to differ significantly in discrimination tests. Although the magnitude of an attribute may not be significantly different, the time or order of perception of that attribute may differ between products. The objective of this study was to collect and analyze the order or time of perception of sensory attributes as a method for improved sensory evaluation of products.

Salad dressing model systems were chosen to deliver the four basic tastes: sweet, sour, salt and bitter, as well as garlic flavor. The same level of each attribute was controlled by using minor changes in formulation. Pairs of products were made that did not different in their profile, but were significantly different in discrimination.

Flavour release was altered through a change in the hydrophilic-lipophilic balance by reduction of oil level in the salad dressings. The reduction of oil in the dressing caused the garlic and bitterness to be perceived earlier. In a similar manner, the increase of viscosity using hydrocolloids created a measurable delay in the release of all attributes.

Twelve panels were conducted using eight trained panellists to test the products both conventionally and using the experimental approach. Statistical analysis demonstrated a significant Attribute by Sample, interaction showing that the sequence truly differed.
These models demonstrated two temporal release patterns: shuffle and shift. Shuffle is the rearrangement of the order of release of each attribute. Shift describes the delay of attributes with no change in their order.

Generalization of this experimental method revealed a large panellist variance due to the complexity of the physical task, making routine application impractical.

Note: This publication is currently unavailable for download.
Categories: Analytical Sensory Methods; Temporal Methods;

Findlay, C.J. The Institute of Food Technologists Annual Meeting + Food Expo 2002. Technical program: Fortification, nutrition, testing and computers. June 16-19, 2002, Anaheim, CA, USA. Scientific Presentation (Oral).

There are three technical issues that inhibit remote sensory analysis data collection. Confirmation of the respondent, security of the data and enforcement of good sensory analysis practice. In April of 2001, Compusense introduced Compusense commuter, a technology designed to deliver Compusense® five sensory questionnaires anywhere and in any language.

The ability to control the presentation of samples and the order and flow of questions was assured by using the original sensory analysis software. The method of keeping the data secure and ensuring the identity of the panellist is relatively simple. The test is encrypted for distribution as an e-mail attachment or download. The test can only be run on computers with commuter Station or Player installed. The respondent must identify themselves with a login and PIN number.

Once the test is completed, the data are transmitted electronically as an encrypted file that is only readable by the originator of the sensory test. Literally any sensory or consumer test can be conducted this way. In May 2001, a Compusense commuter field test was conducted by Mary Gearing, Director of Market Research of Rich’s Products, Buffalo NY, in Shanghai, China. She designed an English-language test which was then translated by Rich's associates in China into Chinese.

The Chinese-language project was put into a commuter presentation and emailed to Shanghai. The Rich's team conducted sensory tests on five products on May 28th. After two days of testing they emailed the commuter files to Compusense, who quickly integrated the result files into the English-language project files. The results were returned to the Rich's team in Shanghai. Ms. Gearing was able to brief Rich's associates with the benefit of statistically analyzed results, complete with cross-tabs, within hours of completing the test. Conventional technology would have required weeks to accomplish the same result.

Note: This publication is currently unavailable for download.
Categories: Computerized Sensory Analysis; Consumer Research;

Duizer, L.M., Bloom, K., Findlay, C.J. 2nd Pangborn Sensory Science Symposium. July 30-August 3, 1995. Davis, CA, USA. Scientific Presentation (Poster).

Dual-attribute time-intensity is a new technique for evaluating the perception of two sensory attributes simultaneously. In this research, the sensitivity of dual-attribute time intensity was assessed in comparison to the single-attribute time intensity test for the evaluation for sweetness and peppermint flavor. Ten trained time-intensity panellists evaluated the peppermint flavored chewing gum samples varying in the intensity of release of sweetness and peppermint flavor.

Testing of the samples was counter balanced so that half of the panellists completed the single-attribute test first while the other half of the panellists completed the dual-attribute test first. For the single-attribute test, the panellists input sweetness perception on a vertically oriented time-intensity line, while peppermint perception was input on a horizontally oriented line using Computerized Sensory Analysis software (CSA; Compusense Inc.).

For the dual-attribute time-intensity test, the CSA program was modified to include the presentation of both a horizontal and a vertical time-intensity scale on the same screen. The panellists were trained to direct the movement of the mouse in two directions simultaneously to represent their perceptions of the two attributes. Comparison of the data collected by both the single- and the dual-attribute time-intensity tests indicated that the dual-attribute test was as sensitive as the single-attribute test.

Note: This publication is currently unavailable for download.
Categories: Analytical Sensory Methods; Computerized Sensory Analysis; Temporal Methods;