The addition of condiments to food is known to affect oral processing behavior, eating rate and acceptance. Although they are a key ingredient in many different culinary traditions, few studies investigated their effects on sensory perception and, in particular, their interaction with carrier foods, which might be crucial to control food acceptance.
The aim of this study was to investigate the interaction/interplay of mayonnaise properties (fat content, viscosity) with carrier properties (hardness of bread and potatoes) by the simultaneous in-vivo real-time analysis of aroma release and perception. Dynamic perception of lemon intensity (Time-Intensity) and in-vivo aroma release of two lemon aroma compounds (limonene, citral) were assessed by 14 trained subjects in triplicate for three mayonnaises consumed alone and in combination with bread and potato of different hardness.
Both mayonnaise fat content and viscosity influence perception and aroma release. Different dynamic release patterns are observed for limonene and citral due to differences in polarity and volatility. Addition of carrier foods (bread, potato) decreases lemon perception intensity of all mayonnaises while, in some cases, it increases the in-nose concentrations. Harder bread decreases lemon intensity to a larger extent than softer bread, whereas only small effects of potato hardness on lemon intensity were observed. The in-vivo release of limonene and citral also depended on the type of mayonnaise. Results are discussed in view of a combination of different mechanisms, including physicochemical (mayonnaise absorbance by carrier, lipid-amylose complexes, bolus partition coefficients), physiological (chewing time and saliva amount) and cognitive effects (presence of one food distracts attention from others).
Both dynamic perception and in vivo aroma release of condiments are affected by condiment properties in a complex interplay with carrier food type and properties. The proposed in-vivo methods allow the dissecting of their complex interaction also in the case of real food.
Lately we see the trend in the FMCG industry to not only measure product performance including overall liking and benefit delivery with the help of explicit but also implicit measurements. Emotions, motivations and attitudes activate and steer behavior without the person often being aware of it. Hence, explicit measurements often fall short in capturing this implicit information.
Implicit measurements that are gaining popularity in the testing of products are those that are based on reaction time (e.g. IAT), or that measure physiological responses such as brain activity (EEG), heart rate, skin conductance or pupil dilation. The downside of the measurement of physiological responses is that we can see the degree of activation but not its direction (positive vs. negative) or even the product’s ability to communicate specific benefits/activate need states.
Furthermore, the authors were interested in the impact of the sensory modality causing emotional response – in our case olfactory vs. gustatory cues.
Hence, a study was set up to investigate the sensory modality impact by testing a variety of citrus scents and citrus flavors applied to a drink using explicit tools (9-point-hedonic scale and CATA for different benefits), a reaction time based method (similar to the IAT but in a more practicable implementation) and EEG.
In this study 40 test persons participated in the EEG part and 100 in the test set-up including reaction time and explicit measurement. Sample presentation and allocation to the three testing set-ups was according an experimental design to minimize carry-over and sequence effects.
Our results show that implicit information can add further insights to explicit measurement with EEG results indicating the degree of activation and the reaction time method adding the layer of need state activation. Furthermore interesting insights on modality-method-interactions were investigated.
The aim of this study was to examine how cross-modal congruence of odour and image valence modulates both neural and perceptual responses to food and non-food odours. Prior research has established that multisensory integration is a fundamental aspect of perception of complex stimuli such as food or wine. A key modulator being the cross-modal congruence of stimulus perceptual characteristics. EEG may provide a unique measure of multisensory integration since it reflects oscillatory neural activity involved in feature binding and integration. Recent research also provides evidence that time-frequency methods can reduce limitations associated with time-locking EEG to odorant stimulation and provide a measure of odour valence, the primary perceptual dimension in olfaction.
To achieve this, 64 channel EEG was acquired while 32 participants rated the pleasantness and intensity of simultaneously presented pairs of affective images (IAPS) and odorants (isopropyl disulfide, vanillin, valeric acid, linalool, eucalyptol and methyl salicylate) diluted in propylene glycol and matched for intensity. For each trial, image and odour presentation was synchronised using a computer-controlled constant-flow air-dilution olfactometer (Lorig, 2000). Odour valence was defined according to each participant’s pleasantness ratings and EEG time-frequency analysis synchronised to the inspiratory peak following each self-initiated trial.
The results showed a significant effect of cross-modal congruence where pleasant odours were rated as significantly less pleasant when paired with unpleasant images, however food and non-food odours did not differ significantly. Odours also induced significantly increased induced EEG power for the 3-6Hz and 10-18Hz ranges relative to a control (air) although cross-modal congruence did not modulate EEG power. Pleasant food odours did show increased EEG power relative to unpleasant and non-food odours. These results are discussed in terms of the correspondence between EEG and affective measures and their relevance for future sensory neuroscientific research in the food and wine industries.
For successful reformulation, it is critical to know sensory difference between original and reformulated products and how such difference changes consumer preference. Comparing to the traditional analytical tests focusing on measuring product sensory difference, affective discrimination tests based on hedonic state of mind has recently been suggested as more ecologically valid test methods to study consumer product discrimination and preference. In addition, consumers’ sensory quality criteria could be influenced by brand information and product concept. Thus, it is required to develop affective sensory test considering brand familiarity to obtain realistic and valid consumers’ response to product changes. In previous study (Park, 2018), the reminder-preference test method that used combination of A-Not A with Reminder (A-Not AR) and 2-AC preference test with the reference-priming by showing TV commercial reflecting product concept and providing brand image has been proposed. This new consumer test method was proved to be more sensitive than the difference-preference test using same-different tests. The objective of this study was to further investigate the performance of reminder-preference test. Lemon-lime-flavored carbonated drink was used as a reference with two types of reformulated samples. Female young adults (N=288) were randomly divided into three groups and performed one of the three tests: A-Not AR test; reminder-preference test; 2-AC preference test. A-Not A test and reminder-preference test were performed with the reference-priming procedure to stabilize consumers’ evaluative criteria, but 2-AC preference test was conducted in blind preference tests. Results computed in d’ indicated that there was no difference in sensory discrimination sensitivity between A-Not AR and reminder-preference test, while the preference sensitivity was higher for reminder-preference than 2-AC preference test. These results suggest that reminder-preference test method taking account for marketing effects could be used as efficient consumer test methods for measuring discrimination and preference at once without cognitive process confusion.
It is widely acknowledged that the traditional approach of asking consumers to rate products for overall liking is a poor predictor of real-life purchase decisions. A few approaches based on consumer choices have been researched in recent years, including the Take Away preferences approach.
As part of a research project on applied consumer method, a study involving 131 UK consumers investigated the added value of using Take Away preferences in CLT settings. Two samples of salty popcorn were assessed in CLT as well as in home placement conditions.
In blind CLT conditions, consumers indicated how much they liked each of the samples and which of the two samples they would buy. In informed CLT conditions, consumers indicated how much they liked each of the samples, whether they would buy each of the samples and at what price, and finally which of the two samples they would choose to buy. At the end of the CLT, consumers were invited to choose two bags of popcorn to take home. They could choose one bag of each sample or two bags of the same. For the two samples taken home, consumers were asked to score overall liking, willingness to purchase and perceived moreishness.
The findings show how much the product packaging/brand (blind vs. informed conditions) and assessment context (home vs. CLT conditions) influenced product liking and purchase intent, and how much liking and buying preferences differ from take-away preferences within a CLT setting.
The findings raise significant questions about the validity of traditional consumer research approaches that Insights and Innovation practitioners use to inform important business decisions. This research demonstrates the benefits of using a Take Away preferences test to simulate real life choice scenarios where respondents make pre-purchase decisions based on product experience as well as exposure to brand and price.
Three commercially available blueberry-flavoured spoonable fermented fresh dairy products (protein-reinforced quarks) were evaluated by a consumer panel in (1) an online survey, (2) during a central location test (CLT), followed by a (3) purchase tracking phase. In addition, (4) implicit product wanting measures were collected during an approach-avoidance task (AAT) using EEG, motor-response kinetics and pupil size responses. The study objective was to compare the power of explicit consumer ratings and implicit product responses to predict product purchase amounts.
134 subjects scored pleasantness, purchase intention, recommendation and emotional responses by emojis for the three quarks after visual exposure to the brand and package (expectation condition) and after subsequent product tasting (perception condition). A random subset of 56 subjects participated in the implicit measurements at the CLT one day after the tasting session. Following this, respondents kept an online purchase diary for one month. Explicit scores of recommendation, purchase intention and pleasantness showed the strongest association with purchase behavior. Of the implicit measurements, ERP-based product wanting showed the strongest association with purchase behavior. Still, explicit responses predicted purchase behavior better than implicit responses. To conclude, if product comparisons are not affected by social desirability of the response, explicit measures predict future purchase behavior better than available implicit measures.
The restoration of context enables novel experimental designs for the investigation of consumer behaviors. In addition to perceptual and hedonic endpoints, other product variables related to consumer behaviors include “consumption occasion (CO)” and “expected price (EP)”. Presently, we assessed whether CO or EP was influenced by increasing levels of context. Sixty-two red-wine consumers evaluated the same 4 red wines in 3 conditions—traditional sensory booths, immersive wine bar, and actual wine bar. For each wine, subjects indicated the likelihood they “would order this wine in a restaurant or wine bar” and “would purchase this wine to drink at home” using a 5-point Likert scale. Additionally, they answered the question “How would you expect this wine to be priced (one bottle, 750 mL)?” by selecting one of the following price ranges: “below $5”; “$5-$10”; “$10-$20”; “$20-$50”; “$50+”. ANOVA results indicate subjects were significantly (p=0.035) more likely to order the wines at a restaurant or wine bar when they were in the actual environment compared to the traditional booth; the immersive wine bar was intermediate. When asked if they would purchase the same wine to drink at home, environment had no significant impact (p=0.064). When estimating wine price, subjects estimated wine prices higher in the actual compared to the traditional booths (p=0.005). Moreover, subjects accurately differentiated price between higher and lower priced wines regardless of environment. Our data indicate that context alters a consumer’s assessment of CO and EP. External cues in the actual wine bar likely raised expectations and positively influenced subjects’ opinions of the wine. However, this effect was only observed when subjects assessed wines for restaurant or wine bar, but not home consumption. This strongly suggests that congruency between the testing environment and questions asked of panelists is important to consider when employing context in product evaluations.
When aiming to differentiate products in a novel sensory direction, there is a risk that consumers who are familiar with the original product would reject the reformulation. First, we need to determine if the sensory direction is a driver of acceptance for the target consumers. Even so, consumers might differ in the extent to which they can perceive product changes in that direction. Consumers can also differ in where their optimum lies for the sensory direction. So, it is important to study such differences in consumer sensory perception through sensory segmentation, which is one of the biggest challenges for successful product and/or market development.
In the current study, 12 bouillon products, varying in mouthfeel thickness were evaluated by 159 consumers for product acceptance using the signal detection-based degree of satisfaction-difference (DOSD) test. Also, consumers sensitivity to differences in mouthfeel thickness were measured by performing a series of 10 2-AFCR discrimination tests on a subset of 4 products.
For sensory segmentation, consumers that performed 8 or more 2-AFCR tests correct were classified as high discriminators, whereas consumer that performed 7 of less tests correct were classified as low-discriminators. The repeated satisfaction responses to the reference product were used to identify four consumer groups: satisfied with reference, reference too thin, reference too thick, and inconsistent results. These results were combined into a matrix summarizing the information into business relevant consumer segments, which could be used to quantify the opportunity of product reformulation. So that estimates could be made of how many people would notice the product change and how it would change their satisfaction.
The segmentation results showed that there was an opportunity to optimize the product into the novel direction as there was a small but sensitive group of consumers that would be more satisfied with a thicker product.
A growing demand for gluten-free products has been observed among both celiac and non-celiac consumers, which makes it interesting to understand consumers’ motivations for consuming gluten-free products. The use of social networks has recently become one of the most popular means for expressing opinions and sharing information about a wide range of topics. In this context, the aim of this study was to analyse the content of messages posted on Twitter about gluten-free products in different food categories.
A total of 16325 tweets in Spanish containing the terms “sin-gluten” and “gluten-free”, generated between September 2017 and January 2018, were collected using the rtweet package from R software. Tweets were processed by automatic cleaning and lemmatization. Subsequently, the obtained dataset was analysed in terms of co-occurrences using VOSviewer software, which creates a semantic network showing conversational connections. Tweets mentioning the four products with the highest frequency of occurrence in the whole dataset were analysed separately: bread, cookie, pizza and beer.
Semantic maps showed related concepts across categories and enabled the identification of the most important characteristics of gluten-free products in each category. Bread, cookies and pizza were more frequently associated with the concepts recipe, flour, cook or hand-made, given that they are products usually prepared at home, using flours from different sources. Beer was more associated with different brands and social contexts. In general, products were connected to positive sensations. In the case of pizza, negative sensations were also identified. Results suggest that the analysis of Twitter data was suitable for obtaining semantic networks that provide insights into what consumers freely express regarding gluten-free products in a micro-blogging platform.
“Overfitting” is a well-known problem in the area of predictive modelling: e.g., a Multiple Regression using more predictor variables than observations will always result in a 100% correct prediction of the dependent variable. However, this perfect fit is a mathematical artefact rather than an empirical finding: it results from “underdetermination”, the situation where too many variables are measured on too low number of observations.
Sensory profiling data is often summarized and analyzed using a similar data format: a matrix with mean scores from a relative lower number of products (in rows) and a larger number of attributes (in columns) is submitted to PCA (“Means-PCA”).
This presentation will discuss the consequences of this underdetermined situation. We’ll also compare it to an application of PCA that is known as “Tucker-1 common loadings model”. The “Tucker-1 PCA” does not only avoid underdetermination, it is also closer to the psychometric application of Factor Analysis: it allows interpreting principal components (or factors) as latent variables, which provide a framework of measuring product perception on underlying sensory dimensions.
Using approx. 400 datasets from the Sensobase database, we can show that the Tucker-1 PCA, in comparison to Means-PCA, provides similar product spaces, but results in higher dimensional factor structures, a better replicability of the factors and easier interpretation of the factor structures.
There is a lack of tools for monitoring individual repeatability, product discrimination and panel alignment in TDS and TCATA. The paper proposes a unified approach, valid for both methods, based on a similarity index between two individual sequences. It amounts to apply the Jaccard index at each time then to average over the full sequence. Statistical inference based on Monte-Carlo simulations and Central-Limit theorem was established. Critical values from these simulations were tabulated for usual numbers of attributes and replicates in order to avoid these simulations to be redone in each application.
Applied to different replicates, the index indicates level of repeatability. Applied to different products, it indicates size of product differences. Applied to different panelists, it indicates level of panelist agreement. Those three indexes averaged respectively over panelists, products or replicates provide panel performances. Further, a discrimination ratio can be computed at both individual and panel levels as the ratio between indexes of product differences and repeatability. If the discrimination ratio is lower than one, than this panelist or this panel seems inconsistent. If not, again simulations can be used for testing significance of the discrimination ratio.
Time discretization requires choosing a realistic size of successive time intervals. For instance, expecting repeatability of attribute citations every second would be illusory. A default value can be the average time between two citations in the study. Ultimately, increasing it gradually until performances are good enough provides an interesting information to the panel leader about the level of reliable temporality the panel is able to offer.
This new way of analyzing TDS and TCATA data are demonstrated in different studies, some of them featuring panelists having done both TDS and TCATA on the same products, and some others in which half of the panel did TDS and the other half TCATA.
Understanding the food preference of animals remains a challenging area in sensory science. An innovative methodological approach, based on the combination of a two‐bowl consumption test and an olfactory discrimination task using a dual‐port olfactometer, was then designed to clarify the drivers of animals' liking. The proposed sensory approach aims to measure the influence of odor on dogs' food selection and to discern a real absence of preference from a lack of olfactory discrimination capacity, which is currently an important flaw of two‐bowl consumption tests results analysis.
Consumption of four pet food products by an expert dog panel was evaluated using two-bowl consumption test. First choice and intake ratio results were strongly linked, suggesting that odor perception influenced the dogs' food consumption. Within this test, two products were equally preferred by the dogs. To provide more depth to the interpretation, we used an olfactometer to measure the dogs' discrimination of the four product odors. A panel of dogs, designated the expert taster panel, was specially trained in olfactory recognition of one of the two equally preferred products.
The main results demonstrated that dogs can discriminate products by olfaction and express the same food preference for different products, even if they have different perceived odors. Moreover, the olfactory experiments allowed us to conclude that the differences in intensity among odors did not affect their discrimination.
Savoury biscuits are among the most popular within the savoury snack category with a wide variety available in the market. Therefore, it is important to evaluate consumer acceptability, emotional and physiological responses to different samples to better understand this broad market. A sensory session using six different biscuits was conducted with a total of N=99 participants (60% Asians and 40% Non-Asians) to assess acceptability (interest, familiarity, texture, appearance and overall liking). The questionnaire was displayed in our Bio-sensory application, which is able to record infrared thermal images to assess body temperature, and videos from consumers to assess the unconscious responses using FaceReader™ for emotions, head orientation and a customised Matlab® code for heart rate. A principal component analysis (PCA) was conducted along with a correlation matrix (p<0.05) for each culture. Furthermore, different machine learning models were developed per culture to assess low and high levels of A) interest, B) familiarity, C) overall liking, and D) texture liking using only the unconscious responses. The PCAs accounted for 58% and 62% of the total data variability for Asians and Non-Asians, respectively. For Asians, there was a negative and significant (p<0.05) correlation between interest and arousal (r=-0.83), and between overall liking and disgusted (r=-0.82). Conversely, there was a positive and significant (p<0.05) correlation between interest and happy (r=0.91), and a negative correlation between overall liking and angry (r=-0.87). The models from Asians had an overall accuracy of A) 94%, B) 75%, C) 91% and D) 95%, while those for Non-Asians presented higher results A) 95%, B) 95%, C) 96% and D) 94%. The proposed models may potentially be used to assess acceptability from Asian and Non-Asian consumers by only recording a video when tasting the products. This may allow fast-screening of different products with a larger number of consumers.