Death of the expert?

Are you a wine expert? No? A food expert, perhaps? No ….really? If, like me, you have a tendency to eat food more than once a day, you really should have been paying attention! But, attending or not, we all have a vast implicit store of knowledge about food, or at least the food that is typical of the culture within which we live. The average consumer can spot a change in one of their regular foods or drinks without very much effort. But what they can’t do is talk very much about that difference, since most of us lack even a basic food vocabulary. And trying to quantify the degree of difference say between two foods is even harder.

For food producers this is a problem. Maintaining the quality of foods can be a challenge when ingredient supplies or storage conditions or packing materials vary. Especially for those foods that are easily influenced by geography or weather or the health of the plant or animal – dairy, wine, olive oil, tea and so on – the expert taster has been a source of information that helps eliminate those samples with faults.

In addition, much of the sensory study of food flavours has been undertaken using experts – specifically trained panellists – whose training is aimed precisely at producing individuals who respond to sensory qualities in an analytical, non-hedonic, and highly sensitive manner [1]. In other words, they are meant to respond entirely unlike the typical consumer. Panellists who are well trained tend to produce data that are reliably reproducible. In contrast, consumers’ data tend to be highly variable, both within (for example, over time) and between individuals. These distinctions are sometimes described as contrasting objective with subjective data, with the implication that the former is intrinsically more desirable for understanding products.

But the trained panel’s apparent advantages come at a cost of reduced ecological validity - that is, application to the real world of consumers. The point of sensory analysis is ultimately to understand the basis of consumers’ decisions about products, which are intimately linked to hedonics. Thus, sensitive trained panellists may describe product differences that are either unperceived by consumers or, even if perceived, then unimportant in terms of their influence on product acceptability. Conversely, minor product differences could potentially exert major effects on food preferences. 

Linking measures of product sensory attributes to consumer preferences implies that trained panellists and consumers actually perceive product attributes in the same way.  But this is not necessarily true. We know that food flavours reflect the integration – or melding together – of taste, olfactory, and tactile information, and this integration means that these individual sensory attributes are melded together. But the whole approach of experts is to analyse, to break apart this synthesis, and with it the integration that produces the overall food flavour. As one example, this is evident when we asked consumers to be analytical and it reduces their liking for the flavour [2]. So, a major issue with consumers is not just that they are inconsistent and variable in their assessments of sensory properties, but that they also ‘see’ flavour as wholes rather than a collection of parts

As recently as the 1990s, it was accepted that the way to find out about consumer preferences was to ask them about liking – which they are clearly easily able to do – and to find out about the product by using trained experts, most often former consumers who’d had the hedonics knocked out of them. The small panels of tasters were experts because of their ability to focus on the minutiae of a product’s sensory qualities in a highly reproducible way. So, after 20, 40 or 80 hours of training, panelists can understand the difference between cohesiveness and fracturability, or between kerosene notes and acetone notes. When faced with such distinctions, the consumer merely looks bemused.

More recently, however, it has been seen as acceptable to ask consumers to distinguish among different products, either via sorting them into groups – which tends not to require a sensory vocabulary – or in the past few years identifying words that apply to whatever is being tasted methods do not require the development of an extensive vocabulary of relevant terms by the consumers themselves. These check-all-that-apply (CATA) techniques allow an insight into consumer perceptions of the similarities and differences between different samples or products.

The biggest change, though, appears to be emerging in the area of simply telling products apart, that is, discrimination. In an earlier discussion (see The highly discriminating consumer), I pointed out the increasing recognition being given to role of emotions in discrimination and how an emotional state might encourage sensitivity.  In turn, this derives from the different ways in which experts and consumers ‘see’ products.

The ability to discriminate between different tastes, flavours or foods is typically seen as a function of perceptual sensitivity derived from detailed analysis of their sensory properties. Hence, trained panels are recommended for such tasks. Conversely, consumers are thought to be relatively insensitive to the presence of, or variations in, sensory qualities, at least when compared to trained panelists, and there has been a reluctance to use consumers to undertake discrimination tests. This practice, however, relies on two assumptions that are increasingly being questioned. 

The first of these is that high levels of sensitivity to variations in sensory qualities is always important. In fact, trained panels may over-estimate the effect of variations on consumer preferences. Thus, measuring the concentration of trichloranisol (TCA; also known as cork taint) at which wine consumers would reject wine produced a much higher value that that given by wine tasters trained to be sensitive to TCA. In effect, wines rejected as tainted by the experts would have been perfectly acceptable to the majority of wine drinkers [3]. 

So, consumers may be entirely appropriate to use in discrimination tasks when making judgments about what other consumers might perceive in products. The question is then how to find differences between products that are actually relevant to consumers? That is, would consumers actually notice the difference and, if they did, would they actually care? In a recently published study, Kim and colleagues [4] used an affective consumer discrimination method. The essence of this test was to perform a standard test of discrimination known as the duo-trio test in which consumers were asked to match one of two unknown stimuli to a reference. In this case, though, the reference was established beforehand as being preferred (affective groups) or no such pre-test was performed (analytical group). They were able to demonstrate that for those consumers who were genuinely “affective discriminators” – that is, they showed a clear preference for the product used as a reference (as compared to non-discriminators) – their ability to match the samples was higher than for an analytical group, who did not undergo this preference pre-test.

Hence, rather than encouraging these consumers to ignore their preferences, actively utilizing these preferences produced superior discrimination. Of course, the presence of sub-groups of ‘affective discriminators’ sounds like unwanted variability of the sort that does not occur with experts. But we should be used to the idea of preference segments. Variability in consumer perceptions and preferences is currently a rapidly increasing area of research (see for example How sweet it is ... or is it?). The key is not to manage-out consumer variability by turning them into experts, but rather measure, identify and interpret the sources of the variability and ask what it tells us about food preference and choices in the population. Eliminating the impact of individual differences means that important variables that have a consistent impact on product perceptions and ultimately acceptability are ignored.

__________________________________________________________________________

1. Prescott, J., J. Hayes, and N. Byrnes, Sensory Science, in The Encyclopedia of Agriculture and Food Systems., N.K.V. Alfen, Editor. 2014, Elsevier. pp. 80 - 101.

2. Prescott, J., S.M. Lee, and K.O. Kim, Analytic approaches to evaluation modify hedonic responses. Food Qual. Pref., 2011. 22(4): p. 391-393.

3. Prescott, J., et al., An estimate of the “consumer rejection threshold” for TCA in white wine. Food Qual. Pref., 2005. 16: p. 345-349.

4. Kim, M.-A., H.-M. Sim, and H.-S. Lee, Affective discrimination methodology: Determination and use of a consumer-relevant sensory difference for food quality maintenance. Food Research International, 2015. 70: p. 47-54.

Fear of frying

In my last post (July, 2014: A Taste of Emotion), I suggested the possibility that responses to tastes were linked to other emotional responses not merely by association to some third, common factor, but rather that variations in taste responsiveness might be the cause of different degrees of disinhibition (defined as inability to delay gratification) or how much we react to emotion-inducing stimuli of all kinds. Speculation is fine but how do we establish that one thing has caused another? Proving causal links between any two phenomena is always difficult, but we can ask what sort of evidence might make us take an idle speculation more seriously, even in the absence of absolute proof.

My own first question was whether we had any evidence that variation in taste preferences – apparently uniform at birth – emerges relatively early. If it does, then this can potentially be an influence on subsequent emotional development. This search sent me to the bowels of my dusty reprint archives (more strictly, the Endnote^tm program on my computer). The answer is that there is such evidence, at least when we consider sweetness preferences.

The strong preference for sweetness seen at birth starts to decline as the infant reaches the second half of their first year, such that some infants become relatively indifferent to sweetened water [1]. What does increase is variability in response to sweetness and a plausible hypothesis is that this is a consequence of differing exposures to a variety of foods, only some of which are sweet. And there is evidence for this. Infants fed sugar water prior to 6 months maintained a high preference for sweetness in water at 12 and 24 months. In effect, at 2 years of age, these infants were “sweet likers” (see: February, 2013: How sweet it is …. Or is it?). In contrast, those never fed sweet water did not show a strong desire to drink sweetened water over plain water [2]. Hence, one consequence of the practice of feeding infants sugar water is that is maintains the same high preference for sweetness that exists at birth into their second and third years.

Secondly, is there any evidence that taste experiences are linked to other emotions? Again, some of the “side effects’ of sweet tastes suggests that there might be. We know, for example, that sweet tastes can pacify an infant when they are receiving a painful hospital procedure [3]. The mechanism is not exactly certain but is likely to involve the activation of reward/pleasure pathways that inhibit pain. The reason that some mothers administer sugar water is presumably to help pacify their distressed infant. Through such pairings, the infant will learn to associate demands with reward – perhaps a precursor of later inability to inhibit immediate gratification.

But taste hedonics is not the only potential early influence on later emotions. One other obvious food-related candidate that may influence emotional development is food neophobia (FN). When seen in children, typically those aged 2 - 6 years, it is interpreted as an adaptive developmental stage that limits ingestion of unfamiliar, and therefore potentially dangerous, items that might be mistaken for food. It is thought to be mediated by the fear that unknown foods might taste unpleasant [4].

FN varies with a number of different factors (age and education, for example), and can be modified by experience with food variety. It also differs between cultures: in one study, consumers in New Zealand were shown to be on average more FN than a similar group from Japan. Dietary variety, or at least an attitude towards dietary variety, would seem to be the most logical reason for such differences. Nevertheless, FN is a strongly heritable trait [5] that persists into adulthood in substantial proportions of the population – up to 50% in some estimates. 

As in children, high FN adults have low dietary variety and show a reluctance to consider new foods – even if they are simply new versions of something already eaten.  They also dislike larger numbers of foods than their low FN neighbours. For the high FN individual, therefore, eating is associated with either boredom with the limited horizons of their diet or anxiety at the prospect of encountering something unfamiliar (or perhaps both) – in either case, food generates a relatively low level of pleasure compared to that experienced by the rest of us.

If basic tastes such as sweetness and bitterness elicit some of the very earliest post-partum emotions, then pleasure in our early years is clearly also heavily reliant on food. Rewards are often food based – the ice-cream or cake treat – and it is in these early years that children start to associate particular foods with other rewarding experiences, such as a trip to the well-known McDougalls Scottish burger ‘restaurant’ (Slogan: “It could be worse!”). Presumably for the highly FN child or adolescent, there is no such thing as being “pleasantly surprised” by food.

But there are many things about the FN experience that are unknown. How does the development of preferences for new foods develop in high FN individuals? Do high FN diets exclude whole categories of foods? One example might be very spicy or bitter foods, the consumption of which requires tolerance for initially unpleasant qualities (remember that FN is mediated by a fear that something with taste unpleasant). Rozin has suggested that the single most important question that can be asked of someone if you wish to know their food preferences is “what is your culture or ethnic group?”. It may be the follow-up question ought to involve measuring FN.

Strong evidence is lacking for an effect of FN beyond the food domain. However, more general personality traits such as sensation seeking are based on a broad enjoyment of novelty as well as “thrill seeking” in a variety of contexts. On another occasion, Rozin proposed that eating chilli might be a form of benign thrill seeking and so the question remains as to whether early failure to thrill seek with foods may predispose the individual to adopt similar approach when faced with novel experiences in general. 

____________________________________________________________________

1. Schwartz, C., S. Issanchou, and S. Nicklaus, Developmental changes in the acceptance of the five basic tastes in the first year of life. Br J Nutr, 2009. 102(9): p. 1375-85.

2. Beauchamp, G.K. and M. Moran, Acceptance  of sweet and salty tastes in 2-year old children. Appetite, 1984. 5: p. 291-305.

3. Blass, E.M. and L.B. Hoffmeyer, Sucrose as an analgesic for newborn infants. Pediatrics, 1991. 87(2): p. 215-218.

4. Pliner, P., M. Pelchat, and M. Grabski, Reduction of neophobia in humans by exposure to novel foods. Appetite, 1993. 20: p. 111-123.

5. Knaapila, A., et al., Food neophobia shows heritable variation in humans. Physiol. Behav., 2007. 91: p. 573-578.

Implicit implications

How do you know what you know? Much of what we do and the decisions that we make are based on knowledge or attitudes that are accessed without conscious effort. Hence, driving a car would be impossible if we continuously needed to think about each step of the process. Experienced doctors are able to make snap decisions by ‘reading’ a patient and their health without going through an explicit mental checklist. It is all part of the process of the automation of perception and reaction that is characteristics of expertise.

Expertise comes in a variety of forms. In one sense, we are all experts regarding the products we use or consume. Food psychologist EP Koster has often made the point that we instantly know when a food company has tampered with our favourite marmalade, even if we are not sure exactly what has changed. Experimentally, too, this was shown in a study in which very small amounts of tastants (sweet, sour, bitter) were added to foods (cream cheese, orange juice, yoghurt) that had been sampled earlier in the day. Asked to compare the new versions with those they had consumed earlier, the participants were especially sensitive to any negative change produced by minimal increases in bitterness, even though it is unlikely that would have been able to articulate what exactly the change was [1]. In other words, they were retrieving implicit rather than explicit information.

In some cases, information is encoded implicitly in the first place, and our behavior and preferences can be shaped by experiences with stimuli that never entered conscious awareness. Thus, the ingenious study of the German recipients of formula milk to which vanilla had been added showed that these experiences were influential in determining a food choice (ketchup with or without a tiny amount of added vanilla) even 30 years later [2].

Consumer research of any kind is aimed at retrieving knowledge that is assumed to be at the consumer’s fingertips. But what if it isn’t? Social psychologists sometimes study phenomena in which an ‘honest’ answer can not be assumed. Thus, if asked questions about our own attitudes to other races, there is enormous internal pressure to conform to cultural norms, even if we know that the information won’t get further than the researcher’s computer. We do not like to admit even to ourselves that we have attitudes that would meet universal disapproval if voiced.

Consumer researchers sometimes have the same concerns, although this is more motivated by the fact that ratings of liking for, say, a new product is a poor predictor of how well that product will actually do in the marketplace. The rather illogical response is sometimes to blame the consumer: they are obviously stubbornly not revealing their true attitudes towards the product. One consequence is the search for “objective” measures of liking such as brain scanning - as though there was a blind alley between “the brain” and the mouth down which honest opinions get sent.

One approach to the problem of retrieving attitudes taken by social psychologists has been the development of the Implicit Association Test (IAT). This is based on the assumption, not that people’s responses are inherently dishonest, but that there may be attitudes that are not explicitly or consciously available or that are difficult to put into words. The technique works by exploiting the fact that responses to things that belong together (words; objects) are faster than when there is a disconnect between them. So, if I flashed up pairs of words on a screen and asked you to instantly press a key, your response to “mother + father” would be quicker than to “mother + sheep”. Now, as the social psychologists do, imagine if I flashed up the name of another race (you decide which!) plus either of the words ‘dumb’ or ‘smart’. Any difference in reaction times between these two words paired with the race would reveal attitudes without ever having to ask a question about racial beliefs.

As one example, I previously (Le topic du jour: Gout qui importe; December, 2012) reported the distinction between French and American consumers in their attitudes towards food tastiness and health, the results of an IAT study that compared responses to the terms “tasty” with “healthy” and “unhealthy”.

The pairings in the IAT do not need to be words but can also be visual images such as labels, people, places or events. For instance, the IAT was recently applied to examining the fit between marketing slogans and their associated products. Pairings of the words “gentle” and “powerful” with images of the bottles of different brands of mouthwash showed that consumers had internalized the marketing messages of each company. Responses to the mouthwash that marketed itself as gentle were fastest when the bottle image was paired with this label than when paired with “powerful”; conversely, the word “powerful” produced the faster reactions times when paired with the bottle of the other brand, again consistent with its marketing [3].

There are indications that measuring implicit responses is being increasingly recognized in sensory consumer science. At the recent Pangborn Symposium in Rio de Janeiro (see: http://www.pangborn2013.com) several presentations focused on measures of implicit reactions to products, including the measurement of emotions – currently one of the current hot topics in sensory/consumers research – using the IAT. Other reported implicit measures included the measurement of facial responses, although the extent to which these might be consciously controlled is an important methodological issue.

One potential next step is to include product tasting with the IAT and assess the extent to which the product’s sensory and functional properties is congruent with the product image and concept. The final plenary talk at the 2013 Pangborn discussed the issue of fit between a product’s conceptual profile (essentially the set of associations and emotions that it evokes) and its sensory properties, noting that the degree of alignment of the two equates to whether or not a product ‘fit-to-brand’. The conclusion was that significant mismatch between a product’s conceptual profile and what it actually delivers to the consumer presents a serious risk that the product will fail [4].  

                                                                                                                                

1. Koster, M.A., J. Prescott, and E.P. Koster, Incidental learning and memory for three basic tastes in food. Chem. Senses, 2004. 29(5): p. 441-453.

2. Haller, R., et al., The influence of early experience with vanillin on food preference later in life Chem. Senses, 1999. 24(4): p. 465-467.

3. Parise, C. and C. Spence, Assessing the associations between brand packaging and brand attributes using an indirect performance measure. Food Qual Pref, 2012. 24: p. 17-23.

4. Thomson, D. The application of conceptual profiling in brand, product and packaging development. 10^th Pangborn Sensory Science Symposium, Rio de Janeiro, Brasil, 11-15^th August, 2013.

Choosing to like or liking to choose?

How do you know what foods you like? Simple. It’s that inner voice, that warm feeling, that ….. subjective positive ‘something’ whenever certain foods are thought about or mentioned or available. Tapping into this subjective state is, at first sight, relatively straightforward. If I want to know which food you like or which of several you prefer, I can ask you.  Sensory and consumer scientists do this all the time.  You can be asked to provide a rating of liking for a food, or be asked to rank foods in order of preference, or simply choose one food out of several to consume. Each of these approaches accesses some aspect of what you are feeling towards the food at that moment. They reflect the subjectivity of your preferences.

 Deriving subjective measures is sometimes seen as a problem, mainly because we often equate objectivity with reliability and hence see subjective responses as somehow inferior and unreliable. This in turn has led to searches for objective measures of subjective states, the most recent of which is brain imaging, typically via fMRI techniques. While it is perfectly possible to show brain correlates of positive and negative emotions associated with foods, there is the rather important question of what is used to validate the brain response. That’s right …. a subjective report. 

Turning to technologies such as MRI is understandable because it is a natural assumption that technology solves problems. Yet, to help us understand likes and dislikes there are some far more basic questions that need to be addressed. Why, for example, when a person provides a high rating of liking for a food in a sensory test does this often fail to predict what they will purchase at a later date? At one level, the answer is obvious. Eating a small piece of confectionary in a test situation – even if it is judged as very delicious (a 10!) – differs from real life eating in that the context is artificial, the amount is unrealistic, the influence of brand is often absent, and you are removed from all those factors that are influential in making choices.  

Studies in the psychology of decision-making have shown repeatedly that situational factors including one’s mood at the time, and a variety of environmental influences (including, for example, the weather), are incorporated into decisions.  These influences are not only often unrelated to underlying preferences, but they also operate largely outside of our conscious awareness. In other words, your selection of Biffo Brand Baked Beans may be as much about the fact that it is cold outside as it is about your likes and dislikes. This is not good news for a food industry wanting to predict consumer behavior from the basis of sensory tests. It doesn’t mean that the sensory tests are useless, merely that they serve only to make sure the sensory properties are within acceptable parameters.

 Unfortunately, there’s more bad news. It is widely assumed that attitudes, beliefs and emotions, including preferences, precede behaviours: I wear my seat belt because I value road safety; I smile because I am happy; I choose food A over food B because I prefer food A. But none of these things are necessarily true, at least not all of the time. Let’s take emotions. More than a century ago, the pioneering psychologist William James wrote about expression as being central to the experience of emotions, and raised the possibility that we might experience subjective emotions as a consequence of monitoring our physiological and behavioural responses to situations. According to James, it wasn’t at all clear that the fear we experience when being chased by a bear (those were dangerous times!) is not due to the fact that we feel our heart racing, and our palms sweating, rather than vice versa. A somewhat similar view was proposed decades later by another psychologist, Stanley Schacter, who showed that whether a drug makes you anxious or happy can depend entirely on the context in which its effects are experienced.

 Most recently, a variant of this idea – the facial feedback hypothesis – was proposed to explain how facial expressions might not only reflect emotions, but also produce them. This hypothesis was tested in an ingenious experiment in which facial muscles were manipulated by having the participants hold a pen between their teeth, parallel to the mouth, forcing the facial muscles into a smile. Compared to a control group who simply held the pen in their hand, these participants rated a set of cartoons as funnier. In effect, forcing a smile produced the same type of mood state usually thought to produce smiles. 

What does all this social psychology have to do with food likes and dislikes? A recent study has confirmed a number of findings from the past decade that the decisions and choices that we make play an active role in producing the preferences that we assume lead to those choices in the first place. Alos-Ferrer and colleagues [1] asked groups of university students to rate their liking for holiday destinations. The experimenters then selected four of these: two were moderately, but equally liked; one was liked less than these; and another was liked more. Each of the moderately liked destinations was then paired with one of the others, and the students were then presented with these pairs and asked to pick one from each pair. This produced a situation where one of the equally liked destinations was paired with a less liked destination (and therefore likely to be picked) and the other was paired with a more liked destination (and therefore unlikely to be picked). In other words, the students were forced into choosing one alternative and rejecting one alternative, when both were previously rated equally liked. When the students were once again asked to rate the destinations, the researchers found that the chosen destination was now preferred over the one that wasn’t chosen.

 One explanation for this effect – now shown in several studies using different types of stimuli – is that increases in liking are based on observations and interpretation of our own past behavior. For example, we might infer from that fact that we have already chosen an alternative that that alternative was previously valued, and this in turn creates a biased view of the alternative’s current value. An account of choices was also provided by Festinger’s cognitive dissonance theory, first proposed in the 1950s. Festinger suggested that a choice between equally valued alternatives creates an unpleasant psychological tension (or dissonance) that is relieved when we decide that one alternative is actually superior to the other. 

Both of these views rely on recall of past choices. However, in another recent report in which choices determined preferences for odours [2], the participants were asked to recall their choices. It was found that preference changes did not rely on an explicit recall of which odour was chosen and which one wasn’t. This suggests strongly that the change is preference was not simply a way of justifying the choices that had been previously made. Another explanation not requiring an explicit memory of choices comes from studies showing that neutral stimuli used as cues become more valued when they signal either the availability of choice [3] or are associated with an evaluation [4]. Thus, choice seems to be intrinsically rewarding and so that anything associated with choice becomes preferred.

 This being the case, should we be relying on choices or subjective measures of food preferences? From the point of view of assessment of consumer likes and dislikes, it is clear that the process of forcing a choice induces a preference. We might expect then that sometimes choices and preferences are highly correlated and sometimes not, depending on how such tests are conducted. But is this any different from what we all experience in a supermarket? Probably not, but recall that choices can be decided by aspects of the context that may be unrelated to the product chosen. Thus, those situational factors that momentarily influence the choices that you make now become highly influential in shaping your longer-term preferences. This does not mean that either choices or measures of preferences are meaningless, just that they are more complex than we generally assume.

                                                                                                           

1. Alos-Ferrer, C., Granic, D-G., Shi, F. & Wagner, A.K. (2012). Choices and preferences: Evidence from implicit choices and response times. Journal of Experimental Social Psychology, in press.

2. Coppin, G., Delplanque, S., Cayeux, I., Porcherot, C. & Sander, D. (2010). I'm No Longer Torn After Choice : How Explicit Choices Implicitly Shape Preferences of Odors. Psychological Science, 21: 489-493.

3. Gast, A. & Rothermund, K. (2011). I Like It Because I Said That I Like It: Evaluative Conditioning Effects Can Be Based on Stimulus-Response Learning. Journal of Experimental Psychology: Animal Behavior Processes,    37: 466-476.

4. Leotti, L. A. & Delgado, M.R. (2011). The Inherent Reward of Choice. Psychological Science, 22: 1310-1318.