This report describes the results of a two-phase research project into people’s understanding of nutrition labelling. It was funded by the Ministry of Agriculture Fisheries and Food. The research was motivated by a recent European Community Directive – the 1990 Nutrition Labelling Rules Directive – which prescribes the format in which nutrition labelling must be presented. The Directive states that nutrition information should be shown numerically, but it allows for the possibility of including graphic representations according to formats yet to be determined.
The aim of the research was to assess whether people could understand numeric nutrition information and whether adding words or graphic information in various formats would help them to choose a healthier diet.
Phase 1 of the project comprised a series of group discussions. Participants were asked about their use of, and preference for, different nutrition labelling formats. Phase 2 consisted of a series of experimental studies in which participants were tested to see which formats were most useful in helping them make decisions about foods.
Results of phase 1: group discussions.
Most participants said they found numeric nutrition labelling difficult to understand because they had insufficient background knowledge to interpret the information given in this form. They said they would prefer an alternative scheme where nutrient levels were expressed using the words ‘high’, ‘medium’ and ‘low’.
When asked to make judgements about single foods using numeric information, participants regularly made inaccurate assessments. This was because they did not know whether the figures represented high or low nutrient levels in relation to medical recommendations. Participants only used numeric information successfully when comparing foods where differences were simple and obvious.
Results of phase 2: experimental studies.
These studies examined how adding words or graphic information affected participants’ speed and accuracy when making judgements about single foods, or when comparing two or more foods.
Verbal banding, where words such as ‘high’, ‘medium’, and ‘low’ were used to indicate nutrient levels was the most consistently helpful format tested. (Examples of formats which were tested are shown in the finished report).
Graphic formats were found to be helpful to some people when they were direct representations of the nutrient levels. In particular supplementing verbal banding with bars where the nutrient level was directly related to the length of the bar helped participanfs who said they did not regularly usenutrition information on food packets.
However graphic formats which attempted to evaluate the nutrient level tended to confuse people. For instance a format which gave two stars for low fat and no stars for low fibre caused people to make errors in their judgements about foods. The participants were confused by the inverse relationship between numbers of stars and the fat content as opposed to the direct relationship between the number of stars and the fibre content.
Dietary reference values (DRVs) indicating the recommended daily intake of each nutrient, and representations of the amount of a DRV supplied by a single serving, both helped some participants in their decision making. This was particularly the case for participants who were more interested in nutrition issues, and when the proportion of the DRV supplied by a serving was presented graphically.
Other findings emerged from this phase of the study. In particular, it was observed that extra information on the packet, such as health and nutrition claims, can affect people’s understanding of nutrition labelling, whatever the format. In addition, people found it difficult to learn how to use more than one format for nutrition labelling.
Conclusion
This study provides evidence for the view that supplementing numeric nutrition information with words or well-designed graphic information can help consumers chose a healthier diet. Banding systems using words or graphic representations are more versatile and generally more helpful than formats based on DRVS.
Phase 1: Report of group discussions
Background to the study
The aim of this, discussion group, phase of the project was to find out what understanding of nutrition and other issues consumers brought to bear on decision-making about food. We took, as a starting point, one of the numerical formats for nutrition labelling that can be presented following the EC Nutrition Labelling Rules Directive* (see Figure 1), and investigated how consumers might use that format. The Directive provides for the presentation of nutrition information in graphic format, in addition to a numeric listing. So supplementary verbal and graphic representations of the numeric nutrition information were devised, and their effectiveness was explored in the context of tasks that consumers are likely to carry out when making decisions about foods to use or buy.
| per 100g | per serving (140g) | |
| Energy Protein Carbohydrate of which sugars Fat of which saturated Dietary fibre Sodium | 1487kj/355kcal 9.4g 22.8g 0.5g 26.0g 11.0g 0.0g 0.7g | 2081kj/497kcal 13.2g 31.9g 0.7g 36.4g 15.4g 0.0g 1.0g |
Figure 1: One of the numeric nutrition information formats that can be Presented following the EC Directive on Nutrition Labelling Rules. Energy is expressed in kilojoules and kilocalories, and seven specific nutrients are listed. The presentation of nutrition information per 100g is required. The additional presentation of nutrition information for an average serving is optional.
We were aware that exploring the use of nutrition information through group discussions was more likely to give a picture of people’s declared comprehension or concern with nutrition issues, rather than their actual comprehension. So we included a series of questionnaire tasks within the discussions. Participants carried out the questionnaire tasks individually and then di,5cussed them as a group. The questionnaire tasks provided us with individual profiles of participants’ comprehension and likely use of nutrition information, which complemented the information gathered in the discussions.
We should emphasise here that discussion groups cannot provide a comprehensive view of the population’s information needs, and we were not attempting to find out the level of comprehension or information needs of particular social classes. When we selected participants for the discussion groups, we aimed to include people for whom nutrition information on food packaging might be helpful – people who were the prime decision-makers about food shopping in their household and who were aware of links between nutrition and coronary heart disease. We also aimed to include people with different levels of awareness of nutrition issues in order to make sure that our conclusions were not biased towards people with either more or less prior information about nutrition and health.
This first phase provided background for the development of sets of numeric, verbal and graphic representations of nutrition information which are tailored to consumers’ understanding and needs. In the second phase these representations were tested with individual consumers, in decision-making tasks that resembled decisions made in shopping or planning meals.
* Commission of the European Communities (1990). Council Directive of 24 September 1990 on nutrition labelling for foodstuffs. 90/496/EEC.
Summary of main findings group discussions
Declared use of nutrition labelling
Most participants in the study claimed they currently looked at nutrition labelling on foods, although some said they looked only occasionally. Participants with both relatively high and standard levels of interest in nutrition and health claimed they looked for information about calories, sugar, fat, salt, fibre (and also additives). The predominant factor looked for was fat content. The more interested participants claimed they looked for several different factors; standard participants usually looked only for a single factor.
Use of weight information in nutrition labelling
Participants found it difficult to estimate what 100g portions of foods looked like. Additionally their estimates of the amounts of foods that would constitute a serving- for themselves differed considerably for some foods. They said they did not use the weight information given in nutrition labelling (some were put off by the use of grams rather than ounces; many felt that manufacturers’ recommended serving sizes were not appropriate). Health considerations usually influenced decisions whether or not to buy or eat a particular food, but not how much of that food to eat.
Declared comprehension of nutrition terms
At least 50% of the interested and standard participants claimed that they fully understood and were influenced in decision-making by the protein, carbohydrate, sugar and fat content of food. In addition, at least 50% of the interested participants also claimed they fully understood and were influenced by energy, kilocalories, saturated fat and dietary fibre. Saturated fat was less well understood and influential than fat. Sodium and kilojoules were poorly understood, and not influential.
Reception comprehension of nutrition terms
Interested and standard participants showed similar, moderate levels of performance in a multiple choice task, despite the interested participants’ declarations of their comprehension of more nutrition terms than standard participants. Performance was best for the terms: fat, saturated fat, saturates and salt. Subsequent discussion of the task showed that even though participants answered the multiple choice questions successfully there was confusion about the relationship between saturated fat and fat, and the role of sodium in a healthy diet.
Production compre ension task
Participants had difficulty explaining carbohydrate/sugar and fat/ saturated fat inclusion relationships, even though many had noticed the typographic signalling of these relationships in nutrition labelling. The interested participants were only slightly better at explaining the relationships than the standard participants.
Participants recommendations for eating for a healthy heart
The recommendations of both interested and standard participants tended to be simple maxims to avoid specific foods or general recommendations for ‘balance’. They did not display particular appreciation of the role of different nutrients in the diet.
Using numeric information to assess single foods
Both interested and standard participants made inaccurate assessments of the levels of nutrients in single foods because they did not know what levels of nutrients might be appropriate to look for.
Using numeric information to compare foods
Participants could make simple comparisons between foods using numeric information, but not complex comparisons where, for example, a higher level of fat in one food had to be offset against a lower level of sugar. They tended to use a single nutrient level (usually fat, probably because of the bias of the discussion groups) as a yardstick to assess the food as a whole.
Attitudes to using numeric nutrition information
Some participants found numeric information difficult to use. Many felt they had insufficient background knowledge to understand the numeric information. Some participants suggested that information about the recommended daily intake of nutrients should be presented as a supplement to help interpret numeric information. Others thought direct health warnings about the levels of particular nutrients would be useful.
Preferences for selected alternatives to numeric information
From a range of formats for the presentation of nutrition information (Figure not includeded here) the preferred format for most decision-making tasks was a banding scheme, used selectively for nutrients felt to have a bearing on health. Some participants preferred a banding scheme for all nutrients, and others preferred a banding scheme supplemented by bars to give an immediate graphic emphasis to the levels of nutrients in foods. Participants felt their understanding of nutrition was sufficient for them not to be confused by different mappings in banding schemes, for example in bandings for fat and fibre, where ‘high’ is bad for fat but good for fibre. Questions were raised about how banding schemes were worked out, and what authority they might have.
Many interested participants gave high ratings to formats showing the percentage of recommended daily allowances provided by a serving of the food. Ratings for percentages of recommended daily allowances were highest for tasks where participants felt there would be plenty of time to make decisions.
Conclusions from group discussions
3.1 Using nutrition information in decision-making
The combined group discussions and questionnaire responses helped us build up a preliminary picture of peoples’ understanding of nutrition labelling, of how that understanding is brought to bear on decisionmaking in shopping and planning meals, and of other information sources that are drawn upon as specific decisions are made.
The results show that people are more likely to draw on their own background knowledge, of nutrition, rather than information presented explicitly on food packaging. This is not at all surprising considering the busy schedules that most people have – many participants in this study emphasised that most of their shopping is done under time pressure and that there is rarely time to make detailed consultations of nutrition information.
The discussion groups showed that much of the background knowledge that people do draw on is vague, especially the knowledge that they claim they use to ensure they have a balanced diet (section 7, on people’s comprehension of and response to nutrition factors in the full publication shows this). Some may have more useful background information in the form of strategies to avoid certain foods that are linked with either heart disease or cancer. However, this strategic information may not be very helpful when they are making decisions about processed foods, such as prepared meals, that are not covered by maxims such as ‘eat less animal fat’.
People do look at explicit information on packaging when they are concerned about following a special diet. In most cases they are looking at single factors, such as calorie content, presence or absence of certain nutrients (such as fat, sugar or salt), or additives. They may also look at the packaging to find out the weight of the contents for price comparisons. If they are looking for more than one factor they may look for information in different places on the packaging.
Generally people appear to process the nutrition information currently available on food packaging at a superficial level. If they are looking at nutrition information they do not involve themselves with the numeric information to any degree, unless they have specific health reasons for doing so. They do not make numerical comparisons between the amounts of nutrients in different foods and, if they make comparisons between the levels of nutrients in single foods, these may well be incorrect because the comparisons may be based on assumptions that quantities of all nutrients can be evaluated on the same scale.
3.2 Users with different levels of interest in nutrition and health
Many consumers choose not to prioritise health considerations when they are making decisions about food. Others feel there is not enough time to think about health issues. Others may be intimidated by the amount and technical appearance of nutrition information presented to them, especially when it is presented in metric units, or may not be comfortable with the kinds of calculations necessary to be able to make use of that information.
Differences in the comprehension for people with different levels of interest in nutrition were not obvious in the reception comprehension tasks described in sub-section 7.2 of the full publication (these kinds of tasks are commonly used to assess people’s comprehension of nutrition issues in questionnaire studies). But differences became apparent in more complex production tasks.
Even though it was possible to isolate a sub-group of consumers who had a more sophisticated understanding of nutrition issues, and might well be interested in more detailed nutrition information than other consumers, the differences between the two sub-groups should not be exaggerated. Many of the more interested participants still had misconceptions and confusions about aspects of nutrition and health.
3.3 Matching nutrition information to tasks and users
Nutrition information could be tailored both to the tasks consumers carry out when making decisions about foods, and to the needs of consumers with different levels of interest in nutrition. It could be an immediately accessible and comprehensible representation of nutrient content, presenting information relating to a range of health issues (consumers following very specific diets might still need to go beyond this immediate presentation to more detailed numeric information). The information would then help consumers place individual foods within the context of their daily requirement for specific nutrients as they plan meals or make decisions about portion sizes.
Immediate information access via a banding system
There were calls for information saying ‘what’s good and what’s bad’ in foods, and for the equivalent of government health warnings on cigarettes, but food manufacturers are unlikely to adopt such a practice. A banding system, either for all nutrients or for nutrients considered relevant to health, would be a more suitable alternative. Such a system would give a ‘first line’ of information, preceding numeric information. Banding systems were tested out in the discussions, and all the systems tested (verbal formats or combined verbal and graphic formats) were popular with all groups, although some participants thought that graphic representation of the banding system was unnecessary.
Recommended daily allowance information for detailed consideration
There were additional calls for contextual information that would help consumers relate the numeric information for individual foods to daily requirements. Systems showing servings in relation to recommended daily allowances (RDAs) (in numeric and graphic formats) were presented to the discussion groups. Although some (more interested) participants saw their value, many found them difficult to understand, and even intimidating. It’s worth bearing in mind that in the groups the RDA information was shown as the main source of information complementing standard numeric displays. It may be that, if it is used in combination with a banding system, which prepares users for the information they may need to interpret, it could be more successful. RDAs would be unlikely to be used by all consumers, but could be available for those who were more interested to draw on when required.
3.4 Following up phase 1 of the project
Our next step was to conduct detailed testing of individual subjects carrying out decision-making tasks using different formats of nutrition information (phase 2 of the project). Following the findings of phase 1 we decided to focus on banding systems and representations of RDAs.
The discussion groups raised some specific questions about the formats used for banding systems: whether they should be selective or comprehensive, whether they should be verbal or combine verbal and graphic elements. Further questions could be asked about how people’s efficiency in using banding systems might be affected by the number of levels of bands in the system – the systems examined in this phase had three bands (high, medium, low), but both two- and four-level systems could be investigated.
There were also questions about how RDAs should be represented: the graphic system used in the descussion groups was not popular, and a suggestion was made for a unit based system (dividing RDAs into ten units). We decided to compare this kind of representation with listings of percentages of RDAs.
For the detail breakdown report for phase 1 (Chapters 4 though 9) see the Printed Edition available from HMSO.
Phase 2: Report of experimental studies
1.1 Reasons for the research
The EC Nutrition Labelling Rules Directive (1990)* regulates the presentation of nutrition information on food packaging, specifying when nutrition information must be given, the nutrients that must be shown, the order in which they should be presented, and their units. It also provides for the possibility of supplementing numeric nutrition information with information in graphical form, according to formats that are to be decided.
The aim of this, second phase of the project was to examine ways of presenting nutrition information that might help consumers make decisions about foods to use or buy. Investigations of different graphic representations formed a large part of the study. However, we did not examine graphic formats in isolation, but how they might be used in conjunction with numeric information. We also examined verbal information (banding words, such as high, medium, low) used to supplement numeric information, already used by at least one retailer to show the levels of nutrients in foods. We were aware that people prefer text to numeric or graphic representations of information
We tested the different information formats in tasks that resembled those people carry out when making decisions about foods: judging single foods, or comparing foods. Most other studies of nutrition information have gathered people’s opinions about different formats, and how easy they might be to use, rather than examining how people use nutrition information. However there is ample evidence in studies of information design that people’s opinions about the effectiveness of different formats do not predict performance. A recent study by Levy and Schucker (1991)t has shown this divergence of preference and performance specifically for nutrition information, in tasks involving the comparisons of foods.
In phase 1 of this study we gathered people’s opinions about different information formats in a pre-test study, using combined discussion group and questionnaire techniques. The first phase had helped to isolate some common problems people encounter when using nutrition information, which we tried to address when preparing the different information formats tested here. The relevant findings of the previous phase and their influence on the second phase are summarised in section 1.2.
* Commission of the European Communities (1990). Council Directive of 24 September 1990 on nutrition labelling for foodstuffs. 90/496/EEC.
t Levy, A.S. and Schucker, R.E. (1991). An experimental evaluation of nutrition label formats: performance and preference. Presentation to National Food Processors Association: Scientific Forum, Chicago, Illinois.
1.2 Relevant findings of preparatory discussion group study
Background knowledge of nutrition issues
Most participants (from the same population as the people tested in the research described here) had a broad understanding of issues underlying nutrition and health: for example, they were likely to know that it was wise to reduce consumption of fat, saturated fats, sugar and sodium, and increase fibre intake. But they were unlikely to be able to give details of the relationship between carbohydrate and sugars, or fat and saturated fats. They felt they were unlikely to be confused by the different mappings of individual nutrients in evaluating foods: for example that high levels of fat, sugar etc. are considered bad for health, but high levels of fibre considered good. The understanding they had was adequate for them to be able to make choices, such as to avoid foods that are identifiable as being high in fat or sugar, if they wished.
Interpretation of numeric information – banding systems
While participants’ general knowledge of nutrition issues helped them make some choices about foods they knew had particularly high or low levels of nutrients, this knowledge did not always help them use numeric nutrition information because they did not know what quantities (for example, grams per 100 grams) represented high or low levels of nutrients. They tended to judge all nutrients on the same scale, which resulted in their making errors, particularly for nutrients, such as saturated fats, fibre, and sodium, where relatively few grams per 100 grams or per serving constituted a high intake. So they needed additional information that interpreted numeric information. Banding systems meet this need, and we decided to examine a range of verbal and graphic banding formats in the present research.
Evaluation of numeric information – (health warning) systems
Some participants thought that representations that evaluated the levels of nutrients on health grounds would be useful (even though they were not confused about the significance for health of different mappings of nutrients). They envisaged systems in which ticks were used for nutrient levels that were considered healthy, and crosses for levels considered unhealthy.Such systems would be difficult to apply effectively (many nutrients would fall into a middle band, where neither ticks nor crosses were appropriate). Consequently, we investigated a different evaluative system where different banding levels were mapped on to appropriate star ratings, so that people could use them to assess foods onthe basis of ‘the more stars the better’.
The need for detail – dietary reference values
Some participants were suspicious of the categorical interpretations of numeric data provided by banding systems, and wanted more detailed interpretations. So we examined systems in which levels of nutrients were listed alongside dietary reference values* or shown as a percentage of the dietary reference value. These systems were more complex, but gave a fuller account of the nutrient levels of foods, than banding systems.
* in the course of this research the Committee on the Medical Aspects of Food Policy published their Dietary Reference Values for Food Energy and Nutrients of the United Kingdom. We have therefore used the term dietary reference value (DRV) in the second phase of this report, where we used recommended daily allowance (RDA) in the first.
Use of graphic representations
Participants were divided about whether graphic representations were desirable or not. Some felt they added unnecessary clutter to complex displays, others felt their ‘visual impact’ would speed up decisionmaking. Whether or not graphic representations are helpful to consumers depends on the quality of the particular representations (see Macdonald-Ross, 1977).* The details of any particular graphic representation are usually a consequence of a range of constraints, including the nature of the information to be represented, current traditions for representing different kinds of information, assumptions about how the user may expect to see information displayed, and the preferences of the people producing the formats. We recorded the decisions made in the design of the graphic formats in order to give as full a background as possible to the materials tested.
1.3 Outline of the methods used in the experimental studies
The tasks
In this second phase, we tested nutrition information presented in different formats, in short decision-making tasks. Participants were asked to judge how wise a choice individual foods might be for a healthy diet, or which of two or three foods might be the wiser choice for a healthy diet. Immediately after they had made their decision they had to give the reasons why they had made that decision. This task contrasted with the task used by Levy and Schucker (1991), whose participants were shown nutrition information for two products at a time, and asked to identify the differences between the two products, only assessing their relative healthiness as a secondary task. Levy and Schucker describe their task as an ‘information search’ task. Information search was a major component of our task, but the search was made in the context of decision-making that resembled real-life decision-making about the relative healthiness of foods.
The nutrition information tested was for real foods that can be bought currently in British shops. In most of the tests, participants were shown the nutrition information on flat sheets of paper. The name of the food was always given with the nutrition information. We felt it was important that people should know what foods they were judging because in real decision-making tasks people’s choices are bound to be influenced by their expectations about individual foods. For example, they may expect breakfast cereals to be a source of fibre, puddings to be high in sugar, and so on. One indication that a particular format for nutrition information was failing to help consumers in their decisionmaking would be if participants tended to base their judgements more on their preconceptions about individual foods than on the information available in that particular format.
We also carried out tests using nutrition information presented on food packages to gain a more complete picture of the process of using nutrition information. We felt it would be likely that people would be influenced by additional information on packages, such as nutrition and health claims, or images of the food, as well as by nutrition information, andtheir prior knowledge of the food.
Additionally in Study 1a we tested use of nutrition information when people had beengiven introductory summaries, explaining the main issues underlying nutrition and health, and introducing the principles of nutrition labelling.
* Macdonald-Ross, M. (1977). How numbers are shown: a review of research on the presentation of quantitative data in texts. Audio-visual communication review, 359-409.
The people tested
We tested women (social class C1) between the ages of 25 and 45 who were the principal decision-makers about shopping, and principal shoppers for their household (see details of recruitment procedure under Participants, in section 4.2 and recruitment questionnaire in Appendix 2 of the full printed edition). All participants had a general understanding of nutrition and health, which they demonstrated by making correct recommendations for healthy eating. But they were then screened further, to separate out individualswith different levels of interest in nutrition issues:a ‘standard’ and an ‘interested’ group. The screening was based on participants’ own declarations about their interest in nutrition issues, and their account of how often they consulted nutrition information. While the screened groups had similar understanding of nutrition issues, it was likely that participants in the interested group were more adept than the standard group at handling the kinds of nutrition information available to consumers. This adeptness may have been a result of using nutrition information more often (because of their higher level of interest), or may have been the cause of their tendency to use nutrition information (if they found it relatively easy and, therefore, rewarding to use nutrition information, their interest in nutrition issues was likely to be sustained).
Measurements gathered
Our main measurements of the effects of the different nutrition information formatswere:
- the length of time participants took to start responding (that is the length of time they took to say whether a food was a wise choice for healthy eating, or which one of two or more foods was the wiser choice)
- the length of time participants took to explain their response
- the number and quality of reasons given in their explanations.
We did not score the decisions that people made because, in many cases, there was no ‘right’ decision. For example,a single food might be high in sugar, but relatively low in fat, and fibre; or, in a comparison, one food might be high in fat, but also high in fibre, another low in fat, high in fibre, but also high in sugar. Individuals’ decisions would depend on the priority they gave to different aspects of nutrition (for example, whether they thought cutting down on sugar was more important than cutting down on fat), and these priorities were not the focus of the study. Our interest was, given any priorities or preconceptions that people might hold, how did the different formats, and the way the formats made information available, influence factors people took intoaccount in their decisions? So we were more interested in the reasons givenin leading to decisions than in the decisions themselves.*
* Note, however, that we made an exception to this strategy of scoring people’s reasons rather than their overall decisions. This exception arose in Study 3a, where there were frequent errors in reasoning with one format, but where it was possible that the errors in reasoning would not affect the overall decision, and so we compared the decisions made with each format.
Summary of main findings of experimental studies
2.1 Study 1
The overall aim of the study was to examine the impact on decisions about foods of supplementing numeric information with verbal banding (high, medium, low).
Study 1a
We compared numeric information with numeric information supplemented by verbal banding, in judgements of single foods, and comparisons of two or three foods. The words high, medium or low were presented alongside numeric information. (Samples of the formats used are shown in Figure 1.) Some participants received explanatory introductions, outlining principles of nutrition and health, before carrying out the judgements tasks, others did not.
| Steak & kidney pie | Steak & kidney pie | ||||||
| per 100g | per serving 142g | per 100g | per serving 142g | ||||
| Energy | 1120 267 | 1635 kJ 390 kcal | Energy | 1120 267 | 1635 kJ 390 kcal | ||
| Protein | 9.8 | 14.0 g | Protein | 9.8 | 14.0 g | medium | |
| Carbohydrate of which Sugars | 25.3 0.2 | 37.0 g 0.3 g | Carbohydrate of which Sugars | 25.3 0.2 | 37.0 g 0.3 g | medium low | |
| Fat of which Saturated | 13.7 5.5 | 20.9 g 7.8 g | Fat of which Saturated | 13.7 5.5 | 20.9 g 7.8 g | high high | |
| Dietary fibre | 0.3 | 0.4 g | Dietary fibre | 0.3 | 0.4 g | low | |
| Sodium | 0.5 | 0.7 g | Sodium | 0.5 | 0.7 g | high |
Figure 1: Numeric information, and numeric information supplemented by verbal banding, tested in Study 1a.
Main findings
For judgements of single foods banding increased the speed of response, and reduced the number of errors made, by eliminating the need for participants to interpret numeric information. For comparisons of foods it had some effects on the speed, and accuracy of reasons, (particularly for standard participants). The results suggested that even when participants had banding information, they still used numeric information, with banding as a back-up. Participants who had been classified as relatively interested in nutrition issues were better able to work between the different sources, using banding to detect large differences, and numeric information for precision.
There was no effect of introductory summaries, and we concluded that their content may have been too complex, and they may have been presented for too short a time, to have had an influence on performance.
Study 1b
We compared numeric information with banding presented without numeric information (a format that would not be allowed by the EC Nutrition Labelling Rules) in judgements of single foods, and comparisons of two foods. (Samples of the formats not shown here).
Main findings
Responses for banding without numeric information were faster than responses for numeric information, and were faster than responses forbanding as a supplement to numeric information, tested in Study 1a. This finding gave support to the conclusion that participants in Study 1a had been using both banding and numeric information when presented with banding.
For judgements of single foods, more reagons were given with bandingonly, because banding words were easier to use than numeric information. For comparisons, banding only reduced the number of reasons given compared to numeric information, because it obscured differences between nutrient quantities that fell within bands.
Study 1c
We compared banding presented alongside numeric information with banding presented separately from numeric information, in comparisons of two foods. The information was presented on food packaging.
Main findings
Displaying banding separately from numeric information increased the time taken to repond, and reduced the number, and accuracy of reason given in responses, compared to banding alongside numeric information. The pattern of results suggested that, ideally, banding should be presented alongside numeric information (although the constraints of packaging design sometimes ruled this out).
The length of time taken to respond with banding alongside numeric information was greater here than in the equivalent condition in Study 1a, where participants saw information on paper. Furthermore participants introduced more reasons based on preconceptions about the foods they were examining here than in Study 1a. These results suggest that use of nutrition information on packages is likely to be influenced by the format of the packages, and by the presence of additional information or images on the packages.
2.2 Study 2
We compared verbal banding with direct mapping of banding into bar charts, in comparisons of two foods. (Note that we presented verbal banding separately from numeric information here, to allow fair comparisons with bar charts, which were also presented separately from numeric information.)
Different numbers of bands were compared: two (high, low) three (high, medium, low), four (high, medium-high, medium-low, low). We compared bandings for all seven nutients in the eight nutrients listed in the EC Directive [ref!] (protein, carbohydrate, sugars, fat, saturated fats, dietary fibre, sodium) with banding for five nutients (sugars, fats, saturated fats, dietary fibre, sodium). (Sample graphs printed in the full-published edition).
Main findings
Bar charts reduced the time participants with a standard level of interest in nutrition issues took to start their responses compared to verbal banding, but had no effect on the number of reasons given. They had no impact on the performance of participants who had been classified as relatively interested in nutrition issues.
The different numbers of bandings did not influence responses, except for interested participants, who responded more quickly and gave fewer reasons in their responses with four-level banding than with two- or three-level banding. Further analysis showed that fewer of the interested participants’ reasons were based on numeric data with fourlevel banding than in other conditions. The findings suggested that interested participants made an early decision that four-level banding could be used to discriminate between foods without recourse to numeric information. Two-level banding was unpopular with all participants because it showed only broad distinctions between foods.
Banding five rather than seven nutrients significantly reduced the time taken to respond, but also reduced the number of reasons given in responses. Participants preferred banding to be displayed for seven nutrients.
2.3 Study 3
Study 3 examined different graphic representations of banding:
- evaluative representations, where, for example, levels of nutrients considered beneficial to health are more, or darker graphic signals, and levels not considered beneficial are signalled by fewer, shorter, or lighter graphic signals.
- direct mapping systems, where, for example, high levels are signalled by more, longer, or darker graphic signals, and low levels by fewer, shorter, or lighter graphic signals.
Study 3a
We compared verbal banding, verbal banding supplemented by a star system evaluating the levels of nutrients, and a star system that evaluated levels of nutrients (without verbal banding). The different information formats were tested in comparisons of two foods.
Main findings
Star systems increased the speed of responses, particularly for standard participants, but also reduced the numbers of reasons given in responses. Furthermore when stars were used without verbal banding, there were more errors in the reasons given (for all participants, but especially standard participants. Errors arose when participants mis takenly interpreted more or fewer stars as meaning more or less of a nutrient (that is, as a direct mapping), rather than a better or worse level of a nutrient (as an evaluative mapping). Note that these participants had not participated in any other studies, and so their errors were not prompted by prior experience of using a direct mapping system.
Study 3b
We compared verbal banding with verbal banding supplemented by bar charts, and verbal banding supplemented by a direct mapping system using shaded boxes. The different information formats were tested in comparisons of two foods.
Main findings
Participants took longer to respond and gave fewer reasons in their responses with bar charts and shaded boxes than with verbal banding. Participants gave fewer reasons in their responses with shaded boxes than with bar charts. These results, which seem to contradict previous results by showing that graphic representations slow down responses, may have been a consequence of transfer effects from Study 3a, which this study followed, with the same participants. The transfer effects illustrate the difficulties that might be caused by having different kinds of systems representing nutrition information operating simultaneously.
2.4 Study 4
We compared different ways of representing nutrition information in relation to dietary reference values (DRVs): a numeric listing of DRVs alongside numeric information; a listing of the percentage of the DRV for each nutrient supplied by a serving of the food; a listing of percentages of DRVs, supplemented by a graphic representation. The different information formats were tested in judgements of single foods, and comparisons of two foods.
In a secondary task, participants were asked to make decisions about foods on the basis of nutrition information presented on packages. The nutrition information included both dietary reference information and banding, and we recorded the choices people made between different sources of information.
Main findings
Participants with a standard level of interest in nutrition issues responded more quickly with graphic representations of percentage DRVs than with numeric listings, or percentage listings, without the poor consequences for the number or accuracy of the reasons given that we had observed in Study 3a. Standard participants performed particularly poorly with percentage listings. Overall, participants who were relatively interestedin nutrition issues took longer to respond and gave more reasons in their responses than standard participants. They were less affected by the different formats of information than standard participants.
In the de-briefing task standard participants tended to use banding information more often than numeric or percentage listings, but showed a preference for graphic representations of percentage listings compared to banding. Interested participants used banding less often, and were more likely to use numeric, percentage, or graphic representations of percentage listings.
Conclusions from experimental studies
3.1 Main themes arising from the research
The use of banding
Banding provides a short cut to the interpretation of numeric information, and so improves judgements of nutrient levels in single foods. It can, however, obscure distinctions between foods in comparisons if the nutrients fall within a banding level. The more fine-grained a banding system, the less likely it is to obscure distinctions. Although participants in this study recognised the value of the detail provided by a four-level banding system (high, medium-high, medium-low, low), many preferred the simplicity of a three-level system (high, medium, low).
Ideally banding should be presented alongside numeric information so that people who want to consult both banding and numeric information can do so easily.
Our finding in favour of banding contrasts with Levy and Schucker’s (1991) finding. However, they tested banding only in comparisons of two foods, and faulted banding on grounds that it obscured distinctions between foods.
The use of direct and evaluative representations of banding
The participants we tested had adequate knowledge of nutrition and health to evaluate direct representations of banding where, for example, high levels of fat and fibre are represented in the same way, despite their different implications for health. A similar basic awareness of basic nutrition issues has been reported elsewhere (Charny and Lewis, 1987; Tate and Cade, 1990).* In our studies we were able to observe participants bringing this knowledge to bear on real decision-making tasks.
In contrast to the success of the direct representation of nutrient levels,we found that banding systems that evaluate levels of nutrients confuse people, so that they make errors in their judgements of foods. Similarfindings are reported by Williams and Poulter (1990) in their trials of amenu labelling scheme.t Evaluative systems have the furtherdrawbackof not being applicable to all nutrients: carbohydrate levels cannot be evaluated since they include both starch and sugars.
* Charny, M. and Lewis, P.A. (1987). Does health knowledge affect eating habits. Health Education journal, 46, pp. 172-6.
Tate, J. and Cade, J. (1990) Public knowledge of dietary fat and coronary heart disease. Health Education journal, 49, 32-5.
t Williams, C. and Poulter, J. (1991). Formative evaluation of a workplace menu labelling scheme. Journal of human nutrition and dietetics, 4, 251-262.
We found further that if direct and evaluative representations of banding are used concurrently, people may not only make errors with evaluative systems, but also make errors with direct representations that they would have been unlikely to make otherwise. Consiquently we have suggested that onlya single kind of representation should be used, preferably direct representation.
The use of graphic representations
As we discussed in section 1.2, it is impossible to give a blanket endorsement or indictment of graphic representations, since their success or failure depends on the kind of representation used. We found that bar charts representing banding, and filled circles representing percentages of DRVs were successful in speeding up the responses of standard participants, and were popular. However, using shaded boxes to represent banding led to some confusions (darker boxes were not always understood as representing higher levels of nutrients), at least in the context in which we presented them (one where participants had already used a different banding system). Although the graphic representations we tested increased the speed of standard participants’ responses, they did not increase the number of reasons they gave in their responses, compared to when they used numeric or verbal information.
Our findings of some success for graphic representations appear to contradict Levy and Schucker’s (1991) finding that bar charts were the least effective of a range of nutrition information formats they tested. But their bar charts were analogue representations of percentage of dietary reference values, and their participants were required to use them for comparisons across all the nutrients in pairs of foods. This is a cognitively demanding task in which the specific graphic format they used seems bound to produce inferior performance to lists of numerals presenting quantities of nutrients for comparison.
Our findings could not be used to support a strong recommendation for the use of graphic representations. But the speed with which they were approached by standard participants suggest that they might be a way of attracting people who would not normally look at nutrition information to do so.
Banding versus dietary reference values
We did not make a direct comparison between these two different ways of representing nutrition information, except in the de-briefing task of Study 4. Standard participants showed a preference for verbal banding to numeric or percentage representations of DRVs, but used graphic represenations of DRVs in preference to verbal banding. The graphic representation of DRVs we used might also be thought of as a ten-level graphic banding scheme. Since it is a direct mapping of nutrient levels, its use would be compatible with our recommendations for direct representations of banding.
It is likely that practical considerations would dictate whether or not DRVs are used: banding appears to be more versatile, since we showed it was effective whether presented verbally or graphically; dietary reference values appear to be most effective when presented graphically, and, consequently, are more limited in their application.
3.2 Putting the use of nutrition information in context
Different formats may suit different users
Throughout our studies, we found standard participants (people who do not regularly use nutrition information on food packages) were more influenced by graphic presentations (bar charts, star systems, or graphic representations of dietary reference values) than interested participants (people who claim they look at nutrition information regularly). Generally, standard participants were less likely to use numeric information than interested participants (Study 4, especially). They had claimed, during recruitment, that they did not have any particulardifficulty using information presented in grams, and there was no evidence to contradict this. However, they appeared to be less able, or willing, than interested participants to navigate different sources of information, or make quick decisions about the most appropriate information to consult.
People with lower numeracy levels than the participants we tested here might find the choice between different sources of information even more difficult to make: they might be drawn to non-numeric representations, but they might find these difficult to understand, too (Levy and Schucker, 1991). The selection of any particular representation of nutrition information will be constrained by the audience being targetted. Formats that help one set of consumers may not translate well to others with different aptitudes for handling numeric, linguistic, or graphic information.
Different formats may suit particular tasks
A further factor that must be taken into account in selecting nutrition information formats is the task that is to be performed with the information. Banding information (whether in verbal or graphic form) improved judgements of single foods by supplying support for interpretations of numeric information (Study 1a). Banding can also improve performance in comparisons of two foods, but at the expense of accuracy in the comparison (Study 1b, Study 2), since if consumers rely on banding information alone, they will miss differences between nutrient levels that fall within a band. However, since we know that people use nutrition information to make judgements about single foods, it may be better to support that task, at the expense of detail in comparisons of foods.
Individual information formats are rarely simply ‘good’ or ‘bad’
Our research has shown that the format of nutrition information, whether numeric, verbal, or graphic can influence people’s decisionmaking about foods. Format influenced performance most obviously in Study3a, where an evaluative star system helped standard participants, in particular, respond more quickly, but also misled them into giving inaccurate responses. Format also influenced performance at a detailed level: moving verbal banding from alongside, to separate from numeric information, in Study 1a, reduced the speed of response, and increased the number of errors made in responding; displaying five banded nutrients, rather than seven, in Study 2, speeded up responses, but also reduced the number of reasonsgiven in responding.
These mixed effects of different information formats demonstrate that decisions about the presentation of nutrition information are not straightforward. It is easy to rule out a system which clearly misled participants. But practical constraints may dictate that supplementary nutrition information is presented separately from numeric information (for example, when graphic representations are used). Furthermore there may be advantages in systems that elicit less detailed responses, but responses that are focussed on nutrition and health (such as a five nutrient banding system), especially if they take less time to use than a broader-based system.