Demonstrating cross-modal enhancement in a real food with a modified ABX test

Highlights

• Taste-aroma interaction between sucrose and vanilla was demonstrated in skim milk.

• A modified ABX matching test was used as a non-scaling test method.

• Compared to a scaling test, larger enhancement effects were found.

• Both congruent and non-congruent aromas showed sweetness enhancement effects.

• Sweetness enhancement was greatest with the congruent aroma in the non-scaling test.

Abstract

Given widespread concerns about sugar intake and health, manufacturers are reformulating their products to help address consumer demand. One common strategy is the use of non-nutritive sweeteners; however, these may come with undesirable bitter and/or metallic side tastes or temporal profiles. One alternative sugar reduction strategy involves leveraging perceptual interactions between taste and smell. Prior work suggests that congruent aromas (appropriate for combination with sugar) can enhance sweetness when added to certain foods. However, demonstrating such enhancement psychophysically is not straight forward, as the conceptual strategy used by assessors can influence whether or not enhancement is observed. Here, a non-scaling approach was used to test for such enhancement. Two simple extensions of the ABX test (namely, matching in ABCX and ABCDX tasks) were used to demonstrate the effect of a congruent (vanilla) and incongruent (beef) aroma on sweetness perception in sweetened milk. Our data suggest both flavors were able to cause enhancement in the flavored milk samples, but enhancement effect was greater for the congruent aroma when compared to the incongruent aroma (p < 0.05). Additionally, greater enhancement was found via the modified ABX matching test, as compared to data obtained via a scaling task.

Introduction

To meet strong consumer demand for sugar-reduced products, food manufacturers are actively seeking novel ways to reduce the amount of sugar in their products while maintaining acceptability (Hutchings et al., 2018, Mintel Group Ltd, 2018). Many of these sugar reduction strategies involve the use of non-nutritive sweeteners (both natural and synthetic). However, in recent Mintel survey data, 40% of consumers indicate they prefer to avoid sugar substitutes due to concerns about their safety, regardless of whether these concerns are accurate or not (Mintel Group Ltd., 2016). Further, some non-nutritive sweeteners have temporal profiles or side tastes that may make them less appealing (Antenucci and Hayes, 2015, Ayya and Lawless, 1992, Azad et al., 2017, DuBois and Prakash, 2012, Reyes et al., 2017, Zorn et al., 2014). At the same time, consumers consistently desire products that taste sweet. These competing demands – i.e., reducing sugar in products and avoiding non-nutritive sweeteners while still making foods that meet desired consumer sweetness levels – leave manufacturers with few options.

Cross-modal interactions, and taste-aroma interactions in particular, may present an alternative or complementary approach to sugar reduction, as some data suggest the addition of specific aromas to a food can potentially enhance perceived sweetness. That is, the presentation of a certain taste and aroma together may enhance the perception of the other modality more than when either is presented independently (Stevenson et al., 1995, Hewson et al., 2008). Examples of complementary odor taste pairs include vanilla with sugar, or lemon with citric acid (Dravnieks, 1985, Noble, 1996, Sakai, 2001). These are commonly considered to be “congruent” pairings, which Schifferstein and Verlegh (1996) defined as the extent to which two stimuli are appropriate for combination in a food product.

Some prior evidence suggests congruency may be necessary for odor taste enhancement to occur. Past examples of congruent and incongruent combinations – e.g., sugar and strawberry odor but not peanut butter odor (Frank & Byram, 1988), soy sauce odor and salt but not sweet taste (Djordjevic, Zatorre, & Jones-Gotman, 2004) – suggest the degree of association between the two is crucial for taste enhancement, which can only occur when they have previously been experienced together. This can be explained via learning of specific combinations through repeated pairing within a naturalistic cultural context or during controlled flavor learning studies in the laboratory (Prescott, 2012, Small et al., 2004, Stevenson et al., 1995, Stevenson et al., 1999, Valentin et al., 2006).

Most prior work on taste-aroma interaction has used model systems (e.g., Arvisenet et al., 2016, Boakes and Hemberger, 2012, Bult et al., 2007, Frank and Byram, 1988, Frank et al., 1989, Frank et al., 1993, Hewson et al., 2008, Jones et al., 2008, Labbe et al., 2007, Lawless and Schlegel, 1984, Lethuaut et al., 2004, Prescott, 1999, Prescott et al., 2004, Stevenson, 2001, Stevenson et al., 1999). Accordingly, it is unclear whether these effects can also be observed in real foods, as studies with real foods are much more limited, although a few exist (e.g., Alcaire et al., 2017, Frank and Byram, 1988, Green et al., 2012, Lavin and Lawless, 1998, Symoneaux et al., 2015, von Sydow et al., 1974).

Within the dairy category, flavored milk (a product that is typically sweetened) represents a highly relevant food matrix for studying taste-aroma interactions: although plain milk consumption has declined, flavored milk consumption has been increasing as flavor has become a more important purchasing factor for consumers (Mintel Group Ltd., 2017). Experimentally, using skim milk (which contains mostly water, ∼9% milk solids, and less than 0.5% fat (FAO, 2013)) helps isolate potential effects of perceptual interactions of taste and aroma. That is, using skim milk preserves high ecological validity of using a real food while also limiting physicochemical effects like volatile release (Rabe, Krings, & Berger, 2003) which could also influence perception.

One of the earliest studies using a combination of vanilla and sugar in milk was reported by Pangborn (1988); however, the focus was on liking and intensity of vanilla milk drinks varying in milk fat and sucrose, and interactions between the ingredients were not explored. Later, Lavin and Lawless (1998) studied vanilla milk (1% milk and vanilla extract at 0.015%); however, they did not include an extended ballot for attributes that would help account for ‘dumping’ that might influence assessment of whether enhancement was observed (discussed further below). More recently, Alcaire et al. (2017) specifically tested taste-aroma interactions for vanilla and sugar in the context of desserts made with skim milk; however, starch levels were varied alongside vanilla and sucrose concentration, making it difficult to draw specific conclusions about cross-modal interaction effects between vanilla and sugar. Thus, despite numerous publications on aroma-taste interactions showing the potential of taste enhancement by smell, important questions remain unanswered.

Previous work indicates that enhancement effects can differ based on the task and the cognitive process participants use to evaluate test stimuli. That is, flavor perception can occur analytically or synthetically, depending on the instructions that are given to participants. An analytical mode focuses on separating the properties of samples into their individual flavor components (taste versus aroma) while a synthetic model requires the stimulus properties to be thought of collectively as a unitary flavor entity (see Lockhead, 1979, Schifferstein and Verlegh, 1996). When an analytical mindset is adopted – which typically occurs during evaluation with intensity or rating scales – an additional complication arises, as the number of relevant attributes to be rated influences whether aroma enhancement of taste occurs (Clark and Lawless, 1994, Frank et al., 1993, van der Klaauw and Frank, 1996). This was first observed by van der Klaauw and Frank (1993) in a study on enhancement of sucrose sweetness by a strawberry aroma. They found that a limited ballot with only a subset of the salient flavor attributes of the product (e.g., a scale for sweetness, but not strawberry flavor) led participants to, in the phrasing later coined by Clark and Lawless (1994), “dump” perceptually similar responses into other attribute scales. That is, a mixture of strawberry aroma and sucrose is rated as being sweeter than the same sucrose concentration without any added strawberry aroma, resulting in what initially appears to be enhancement of taste by smell. If however, participants are given a ballot with sufficient attributes (e.g., scales for sweetness and strawberry flavor), no increase or enhancement is observed. van der Klaauw and Frank (1996) hypothesized that providing a full range of appropriate responses encourages participants to separate the attributes of a sample analytically, while when give a ballot with limited responses, participants are prone to integrating these dimensions (i.e., using a synthetic mode), resulting in apparent taste enhancement. However, it remains unclear if this is true enhancement, facilitated or encouraged by the cognitive mode (i.e., synthetic versus analytic), or rather if it is merely an artifact in translating the perceptual experience into a response due to the demand characteristic of the task. Subsequently, others have observed that aroma enhancement of taste is either weak or absent when an extended ballot is used (Fujimaru and Lim, 2013, Green et al., 2012, Lim et al., 2014, Linscott and Lim, 2016, Wang et al., 2018). Although dumping may potentially explain apparent taste enhancements that occur when intensity rating scales are used, the analytical strategy demanded by scaling may not be representative of what is experienced during a real eating situation. When eating, initial responses to foods are typically holistic and hedonic (Prescott, 1999), where sensory information is integrated into one unitary percept, consistent with a synthetic mindset (Prescott, 1999). This idea that gustatory mixtures are synthetic – the mixing of stimuli forms a unique flavor on its own – has been suggested by several authors (Kuznicki and Ashbaugh, 1979, Kuznicki and Ashbaugh, 1982) and aligns with the view that participants tend to perceive mixtures as a whole, rather than analytically (Schifferstein & Frijters, 1990).

Accordingly, using a direct scaling approach that induces an analytical approach may not be the most useful way to assess aroma-taste interactions. Rather, a scale-free matching task may be an appropriate alternative (McBride, 1983). To date, we are only aware of two aroma-taste interaction studies that have used methods that encourage the use of a synthetic strategy by not using scales. In the early 1990s, van der Klaauw and Frank (1994) asked participants to match strawberry aroma/sucrose or coffee aroma/sucrose mixtures to varying sucrose solutions that came closest in sweetness to the mixture. Lawless and Schlegel (1984) compared direct and indirect scaling of sucrose and citral solutions and found evidence for interactions when using indirect (i.e., non-scaling) methods. In this study, no specific attribute was identified, so participants presumably integrated the stimuli as unitary percepts rather than treating the mixtures as analyzable sets of attributes (Kuznicki, Hayward, & Schultz, 1983).

Here, we built upon the matching approach of van der Klaauw and Frank to mitigate possible response biases that may arise from using rating scales, but modified the task to mimic an ABX test. The ABX test (Munson & Gardner, 1950) involves the matching of a sample (X) to either reference A or B, where one of the labeled references is the same stimulus as the unknown sample X. Further, in contrast to the matching approach used by van der Klaauw and Frank (1994), and in line with the approach of Lawless and Schlegel (1984), we did not direct the participant to any specific attribute, which may also promote a synthetic mindset. Further, we included several sucrose concentrations as references A, B, etc., and asked participants to match a sucrose-vanilla mixture X to the most appropriate sucrose reference. Thus, participants would be encouraged to use a synthetic mindset when conducting the matching test, and we hypothesized this would allow us to demonstrate enhancement with a non-scaling task, thereby avoiding potential artifacts induced by an analytical strategy.

Our modified ABX differed from a conventional ABX test in that in some instances, our sample X would not have an exact match, as none of the sucrose references had any added vanilla. We hypothesized that any enhancement of sweetness by aroma would result in a match with a reference containing a higher sucrose concentration at greater than chance probability. Subsequently, we used the same approach to assess whether congruency between the taste and odor is a necessary condition for enhancement; we did so by comparing a congruent aroma to an incongruent one when paired with sucrose. We hypothesized that a congruent aroma would lead to a greater enhancement than an incongruent aroma.

This manuscript outlines two experiments that attempt to answer the following research questions:

• Can aroma enhancement of sweetness be demonstrated via a non-scaling method that presumably encourages a synthetic mindset?

• Given a range of sucrose concentrations above and below the actual sugar concentration, can a sweetness enhancement be induced by a congruent aroma?

• Does congruency affect the occurrence of aroma enhancement of sweetness?