The authors have declared that no competing interests exist.
Conceived and designed the experiments: SB JNL. Performed the experiments: SB JNL. Analyzed the data: SB JNL. Contributed reagents/materials/analysis tools: SB JNL. Wrote the paper: SB JNL.
The desire to consume high volumes of fat is thought to originate from an evolutionary pressure to hoard calories, and fat is among the few energy sources that we can store over a longer time period. From an ecological perspective, however, it would be beneficial to detect fat from a distance, before ingesting it. Previous results indicate that humans detect high concentrations of fatty acids by their odor. More important though, would be the ability to detect fat content in real food products. In a series of three sequential experiments, using study populations from different cultures, we demonstrated that individuals are able to reliably detect fat content of food via odors alone. Over all three experiments, results clearly demonstrated that humans were able to detect minute differences between milk samples with varying grades of fat, even when embedded within a milk odor. Moreover, we found no relation between this performance and either BMI or dairy consumption, thereby suggesting that this is not a learned ability or dependent on nutritional traits. We argue that our findings that humans can detect the fat content of food via odors may open up new and innovative future paths towards a general reduction in our fat intake, and future studies should focus on determining the components in milk responsible for this effect.
In many Western diets, up to 40% of daily caloric intake is in the form of lipids, despite the fact that the recommended level for most individuals is at least 10% lower
From an ecological perspective, however, it would be clearly advantageous to detect the fat content of food from a distance in order to maximize the chances of finding a source of calories, and before a potentially toxic substance enters the mouth for nutrient evaluation. Support for this basic assumption already comes from animal literature in which the role of olfaction in the preference for fat foods and triglycerides has been clearly established
In three behavioral experiments, we aimed to determine whether humans can detect fat based on the sense of smell alone, using a more natural setting, namely milk samples containing varying ecologically relevant percentages of fat. In Experiment 1, we determined whether humans are able to discriminate between fat content in a natural product based on their sense of smell alone in a North American sample (USA). In Experiment 2, we replicated the experiment in a different population (the Netherlands) where the average daily consumption per capita of milk as a beverage exceeds the US daily consumption by 31%
All participants provided written informed consent prior to participation and all aspects of the study were approved by the Institutional Review Board at the University of Pennsylvania.
Overall, participants were able to discriminate between the different milk samples (mean total percentage correct = 48.6, t(29) = 6.27, p <.001) significantly better than expected chance level (33.3%). Moreover, subjects were significantly able to discriminate the skim milk from medium milk (mean percentage correct = 45.9, t(29) = 3.47, p = .002), as well as skim milk from fat milk (mean percentage correct = 62.6, t(29) = 6.50, p <.001). However, participants were not able to discriminate the medium milk from fat milk (mean percentage correct = 37.4, t(29) = 1.10, p = .282), see
In all graphs: Error bars denote standard deviation, and stars above bar denotes results significantly different from expected chance performance (33.3%, p <.05). S = skim milk, M = medium milk, F = fat milk. See Methods section for further details regarding fat percentage. SM = discriminating between skim and medium milk; MF = discriminating between medium and fat milk; SF = discriminating between skim and fat milk. Dotted line in panel A, B, & C indicates expected chance performance (33.3%).
Participants rated the intensity and pleasantness of the three milk samples significantly different. With increasing fat content, the sample was rated as more intense and less pleasant (intensity ratings: F[2,56] = 35.36, p <.001: mean rating ± SD skim 16.7±1.4, medium 34.8±20.5, fat 47.0±24.3; pleasantness ratings: F[2,56] = 3.90, p = .026: mean rating ± SD skim 51.5±14.8, medium 47.3±18.8, fat 44.0±24.4).
As demonstrated above for Experiment 1, participants were overall able to discriminate between the different milk samples significantly better than chance (mean total percentage correct = 53.1, t(17) = 5.18, p <.001, chance level 33.3%). Moreover, participants were significantly able to discriminate the skim milk from medium milk (mean percentage correct = 57.4, t(17) = 3.31, p = .004), as well as skim milk from fat milk (mean percentage correct = 64.8, t(17) = 5.87, p <.001). Similar to Experiment 1, participants were not able to discriminate the medium milk from fat milk (mean percentage correct = 37.0, t(17) = .73, p = .477), see
Contrary to the finding in Experiment 1, participants did not rate the pleasantness of the three milk samples significantly different (F[2,34] = 1.58, p = .221): mean rating ± SD skim 39.3±23.1, medium 39.8±20.7, fat 32.9±15.1.
In a direct replication of the findings in Experiment 1 and 2, participants were able to discriminate between the different milk samples significantly better than chance (mean total percentage correct = 42.1, t(59) = 4.49, p <.001; excluding BMI outlier, mean total percentage correct = 41.9, t(58) = 4.36, p <.001; chance level 33.3%). Moreover, participants were significantly able to discriminate the medium milk from fat milk (mean percentage correct = 46.7, t(59) = 4.40, p <.001; excluding BMI outlier, mean percentage correct = 46.0, t(58) = 4.21, p <.001), as well as skim milk from fat milk (mean percentage correct = 43.9, t(59) = 3.51, p = .001; excluding BMI outlier, mean percentage correct = 44.4, t(58) = 3.65, p = .001), but not the skim milk from medium milk (mean percentage correct = 35.6, t(59) = .82, p = .417; excluding BMI outlier, mean percentage correct = 35.3, t(58) = .72, p = .475).
There was no significant difference between normal-weight and overweight participants in their olfactory fat discrimination performance; neither on the combined score of all samples (mean total percentage correct for normal-weight 42.8, for overweight 41.3, t(58) = .38, p = .707; excluding BMI outlier, mean total percentage correct for overweight 41.0, t(57) = .45, p = .707), nor for the separate combinations (F[2,116] = .992, p = .677; excluding BMI outlier, F[2,114] = .315, p = .730;
Participants rated the intensity and pleasantness of the three milk samples significantly different (intensity ratings: F[2,114] = 4.45, p = .014; pleasantness ratings: F[2,114] = 6.73, p = .002; see
Ratings (100 mm VAS) ± SD | All | Normal-weight | Overweight | |
Intensity | Skim | 28.0±20.3 | 31.4±21.5 | 24.4±18.6 |
Medium | 26.1±21.2 | 29.8±20.5 | 22.2±21.6 | |
Fat | 33.7±18.6 | 39.8±18.2 | 27.3±17.1 | |
Pleasantness | Skim | 43.8±17.4 | 46.9±14.5 | 40.6±19.8 |
Medium | 44.1±15.5 | 45.7±14.5 | 42.4±16.6 | |
Fat | 37.0±19.0 | 36.1±20.3 | 37.9±18.0 |
There was no significant correlation between either total amount of milk consumed per day (in grams and kcal) or amount of fat from dairy consumed per day (in grams,
In the present study, we aimed to establish the ability of olfactory fat detection in humans using an ecologically relevant setting (milk samples containing varying realistic percentages of fat). In a series of three behavioral experiments, using study populations from different cultures, we demonstrated for the first time that humans can discriminate between varying concentrations of fat in a food product, using only their sense of smell. Moreover, this ability was not related to differences in BMI or dairy fat consumption.
Previous studies have demonstrated that humans can discriminate high concentrations of long-chain fatty acids in vapor phase both retronasally and orthonasally
For all three experiments, our data demonstrate that participants are significantly able to discriminate between the different milk samples based on the overall scores. Specifically, in all studies, discrimination between skim and fat milk was significantly above chance level, which corresponds to the largest difference in the percentage of fat. However, in Experiment 1 and 2, participants were unable to discriminate medium milk from fat milk and in Experiment 3, they were unable to discriminate skim milk from medium milk. Although the absolute difference in fat percentage between skim-medium, and medium-fat was the same (each differing about 1.24%), it might be the relative difference in fat content that is relevant for humans to detect and hence be able to perceptually discriminate between. The difference in fat percentage from skim to medium is a 10-fold increase, whereas the fat percentage only doubles from medium to fat milk. However, another plausible explanation is that the just-noticeable-difference (JND) for these volatiles is in the 1.24% range, thus rendering discrimination between the medium fat concentration and the two endpoints inherently difficult to perform. Future studies should aim to assess discrimination performance between larger ranges of fat contents to establish the JND for fat odor discrimination.
Interestingly, there was no statistical difference in the ability to discriminate between varying fat percentages between different populations (US and the Netherlands) differing in their milk consumption, as well as no individual correlation with either BMI or dairy (fat) consumption. This is in contrast to previous studies by Stewart and colleagues who demonstrated that fat taste sensitivity is associated with both lower fat intake and BMI
In Experiment 1, the intensity and pleasantness of the three milk samples were perceived significantly different; the samples were rated as more intense and less pleasant with increasing fat content. This was not, however, the case for Experiment 2 and 3. The minute differences in vitamin content that existed between the fat and skim milk sample in Experiment 1 might be a factor here. However, even if the small differences in vitamin content mediated the difference in perceptual ratings in Experiment 1, we do not believe that those differences accounted for the main discrimination results; both Experiment 2 and 3 yielded nearly identical discrimination performance results using different milk samples. Our result that the milk samples could be discriminated from each other does indicate there are perceptual differences between them. We hypothesize that perhaps the perceptual differences are not directly attributable to their intensity or pleasantness ratings, but rather originate from an inability to categorize a separate perceptual quality such as ‘creaminess’ that we did not ask for in our experiments. Limitations in alternatives that are perceived as relevant can lead participants to ‘dump’ their response into the categories that are available for them. This phenomenon of so-called halo dumping is well established and might explain the conflicting outcomes between experiments in respect of the perceptual ratings
Dietary fats, or triglycerides, are not known to be volatile. It is therefore unlikely that it is the fat per se in the milk that is the direct odor source which is detected and discriminated; hence it is as yet undetermined what chemical signal the participants are picking up with their nose in the milk sample. However, triglycerides can act as reservoirs of volatile flavor, that is, they can act as carrier of other volatile components that humans can detect in a fat-dependent manner. Alternatively, they are known to interact with other ingredients in the milk, thereby altering the olfactory percept
In conclusion, these data demonstrate that humans are able to discriminate between varying grades of fat, even when embedded within a milk odor. Interestingly, and in contrast to fat taste, this ability is not related to either BMI or dairy fat consumption, suggesting this is not a learned ability or dependent on nutritional traits. The demonstration that humans have a functional olfactory system specific for detecting levels of fat content warrant further explorations into this mechanism given its potential to aid in a general reduction of our fat intake.
(DOCX)
We would like to thank FrieslandCampina for providing us with the milk powder samples, Dr. Kathrin Ohla, Ms. Eva Alden, Ms. Kristen Gregory, Ms. Rosanna Kapel, and Ms. Yfke de Vries for assistance with data collection, Dr. Jeanne de Vries for her help with constructing the dairy consumption questionnaire, and Drs. Mark Friedman and Paul Wise for their comments on earlier versions of this manuscript.