Sensory-Based Action Threshold for Multicolored Asian Lady Beetle-Related Taint in Winegrapes

Insects.

Harmonia axyridis adults were collected during August 2005 in soybean fields at the Rosemount Research and Outreach Center, University of Minnesota, Rosemount. Following collection, adults were held in 1.96-L plastic dishes with ~100 beetles per dish, and maintained at 10 ± 1°C and a photoperiod of 16:8 (L: D) hr. Three days before use, the dishes containing H. axyridis were warmed to 25 ± 1°C with a photoperiod of 16:8 (L:D) hr. Harmonia axyridis adults were provided an ad libitum supply of live soybean aphids, Aphis glycines Matsumara, pea aphids, Acyrthosiphon pisum (Harris), and water in cotton balls placed in plastic Petri dishes (60 mm x 15 mm).

Grapes.

In September 2005, 147 kg of Frontenac winegrapes were harvested from a commercial vineyard considered free from H. axyridis in Lake City, MN. The grapes were hand-picked and 7 kg of grapes were placed into a 22-L polycarbonate container and transported to the Enology Laboratory at the Horticultural Research Center, University of Minnesota, Chaska, where all clusters were carefully hand-sorted and any H. axyridis or other lady beetles were removed.

Infestation to determine reflex bleeding effects.

Artificial infestations of the grapes with H. axyridis were prepared by adding 3.0 H. axyridis/kg grapes to 22-L polycarbonate containers containing 7 kg of grapes. Three container replicates of each treatment were prepared. The containers were closed, inverted, and rolled for 45 sec to mimic the disturbance that might be expected during grape harvest. Beetles were then removed from the containers and grapes were processed and winemaking was started using standard microvinification techniques. Wine was processed separately for each replicate. The effects of ref lex bleeding in the wine taste were compared to the control (i.e., no beetles or reflex bleeding) using a chi-square test.

Infestation to determine sensory threshold.

Artificial infestations of the grapes with H. axyridis were prepared by adding various densities of beetles to 22-L polycarbonate containers containing 7 kg of grapes (Table 1⇓). Three container replicates of each treatment were prepared. The containers were closed, inverted, and rolled for 45 sec to mimic the disturbance that might be expected during harvest. The grapes with H. axyridis adults were then processed and winemaking was started using standard microvinification techniques. Wine was processed separately for each replicate. During the racking process H. axyridis were removed from the treatments.

Wine processing.

The wine was prepared using microvinification procedures. There were 21 batches of wine; one for each replicate treatment combination (Table 1⇑). Crushing and destemming was done mechanically in each 22-L polycarbonate container. Fermentation was conducted at 21°C until dryness in containers approximately 2/3 full. All wine batches were inoculated with Pasteur Red yeast (Red Star, Milwaukee, WI) at a rate of 265 mg/L. After 5 days, wine was racked to glass carboys to finish fermentation. Malolactic culture (CH 35; Chr. Hansen, Hørsholm, Denmark) was added toward the end of fermentation to reduce acidity. After malolactic fermentation completion, sulfur dioxide was added at 50 mg/L. Wine batches were racked to full containers and cold stabilized for two weeks. Each replicate was assessed for pH using an Accumet pH meter (Fisher Scientific, Pittsburgh, PA), sugar content using refractometer, and titratable acidity using standard NaOH titration. Finally, the wine was bottled in standard 350-mL splits with natural corks and stored at 21°C.

Tasting panel.

Panelists (n = 11) were recruited from the University of Minnesota, St. Paul. They evaluated the wines in individual booths, during four sessions, from 1900 to 2100 hr, at the Sensory Center, Department of Food Science and Nutrition. Panelists consumed red wine at least once per month. In a preliminary screening test 10 of the panelists correctly identified the odd sample in three of five triangle tests (i.e., comparing two very similar red wines), while one panelist correctly identified two of five triangle tests.

A series of three-alternative forced choice (3-AFC) tests (Meilgaard et al. 1999) were used for sensory threshold determination. Panelists participated in four sessions, two sets of two sessions one week apart. Samples were randomly assigned to either the first or second week. The serving order of the samples was randomized within each session. After use, each wine bottle had its headspace filled with nitrogen and was refrigerated between replicate days.

At each session panelists were first served “warm-up” samples (O’Mahony et al. 1988) consisting of 30 mL of the control wine and 30 mL of the wine from the treatment with 8.0 H. axyridis/kg grape. The samples were labeled as “no taint” and “tainted,” and panelists were asked to taste the samples back and forth five times or until they could perceive the difference.

After warm up, panelists received the 3-AFC tests. For each test, they were given two samples of the control and one of a tainted sample (15 mL of each sample in a wineglass). The instructions for each test were to evaluate each of the three samples in each set and identify the sample containing the most taint, taking into consideration all characteristics of the samples (smell, taste, texture, etc.). Panelists were asked to expectorate the samples and thoroughly rinse their mouths with water between all warm-up and the test samples. During the first session, panelists were given three 3-AFC tests, one each from the treatments that were infested with 0.3, 1.0, and 3.0 H. axyridis/kg grape. The latter treatment had beetles removed before crushing and destemming. The second session was a replicate of the first session. For the third session, the panelists were given three 3-AFC tests from the treatments that were infested with 0.5, 3.0, and 8.0 H. axyridis/kg grape. (The 3.0 H. axyridis/kg grape tested in this session did not have the beetles removed before crushing and destemming.) The fourth session was a replicate of the third session.

Threshold determination.

A correct response was recorded if the panelist correctly identified the tainted sample in both replicates. In a single 3-AFC test, the probability of correctly identifying the tainted sample would be 1/3 by chance alone and 1/9, or 11%, when the judge had to identify the correct sample in two different randomized sets (replicates).

Logistic regression (proc logistic; SAS software) (SAS Institute, Cary, NY) was used to relate the proportion of correct 3-AFC tests to infestation level. In logistic regression, ln[p/1−p] = α + βx, p was the percentage of subjects who got the test correct, including those who got it correct by chance, x was the number of H. axyridis per kg of grapes processed in making the wine, α was the intercept, and β was the slope. The probability of correct answers, P(x), for a given x, number of H. axyridis per kg of grapes, was estimated solving the logit probability function, P(x) = e^{(α + βx)}/(1+e^{(α + βx)}). The e in the equation is the base for natural logarithms (approximately equal to 2.71828). The 95% confidence intervals of the predicted correct answers were determined using:

where P(x) is the estimated logit probability function for a given x (number of H. axyridis per kg of grapes); z is the value of probability for a normal distribution for a given a (significance level); and SE is the standard error, which is determined by the following equation:

where s_α² is the estimated variance of α (i.e., intercept), s_β² is the estimated variance of β (i.e., slope), and cov_α,β is the estimated covariance of α and β. Finally, we solve the logit probability functions, P(x₁) = e^(x1)/(1+e^(x1)) and P(x₂) = e^(x2) /(1+e^(x2)), where x₁ and x₂ are the lower and upper limits of the confidence interval, respectively (Agresti 2002).

The logistic regression equally was also used to estimate the infestation level (i.e., number of H. axyridis per kg of grapes) and 95% confidence intervals for a given probability of correct answers (Stokes et al. 2000). The predicted infestation levels were estimated by solving the logistic model for a given P, and its 95% confidence interval was estimated as:

where x(P) is the estimated number of H. axyridis per kg of grapes for a given P (probability of correct answers); and SE is the standard error:

Using this method, we estimated four infestation levels of H. axyridis by solving the logistic regression for p = 0.12, 0.20, 0.33, and 0.55. These are the values when 1, 10, 25, and 50% of the subjects got the test correct, without guessing, and the remaining panelists guessed and were correct 11% of the time.

Historically, thresholds are defined as “a stimulus intensity that will produce a response in half the population” (Bi and Ennis 1998). For sensory threshold determination, the 50% of correct answers is taken in addition to the percentage of correct decisions that occur by chance (Meilgaard 1991). By this definition, the sensory threshold would be calculated by solving the regression equation for concentration when p is set to 0.67 for a 3-AFC test with 33% of correct answers by chance (Meilgaard 1991). However, sensory threshold can be used to estimate thresholds of interest other than the ones determined by the 50% of correct answers above chance. For example, thresholds can be determined where individuals can detect off-flavor or taint 90% of the time or off-flavor and taint can be detected by 1% of a population (Meilgaard 1991, Bi and Ennis 1998).