Marmite: checking whether it really is a love or hate relationship


What do you get for the person who has everything? And who you also hate? By Gilda from London, UK (Marmite pop-up shop Uploaded by Edward) [CC BY-SA 2.0 (, via Wikimedia Commons

Jokes about Marmite; most people don’t have strong responses to them. This is unlike the recent news that as a result of potential Marmite price rises, one supermarket might have stopped stocking it. It was generally reported that people were furious with rage, which continued when the dispute was resolved approximately 24 hours later. And because it was opinions on the internet, people said that those opinions were wrong. And because it was definitely opinions on the internet, people went out of there way to say how little they cared about the issue. Whatever your thoughts regarding this particular spread, it’s difficult to deny that the specifics of its one “you either love it or hate it” advertising slogan have been pervasive. So much so that the name ‘Marmite’ is almost synonymous with something which polarises opinion. It’s a real Marmite situation. But what’s the question at the end of the first paragraph that reveals what the rest of the blog post is about? And is it true that people either love or hate Marmite, with no place for yeasty apathy? Luckily, surveys, maths and toast could be used to check.

The information regarding people’s opinion on Marmite was taken from the YouGov UK website. According to this website, YouGov survey approximately 5 million online panellists from across 38 countries including, among others, the UK, USA, Denmark, Saudi Arabia and Europe and China. They claim that their panellists are from a wide variety of ages and socio-economic groups, allowing them to create online samples which are nationally representative. The UK panel, from which the data used here were taken, includes more than 800,000 people. So essentially I went to the YouGov UK website, searched for ‘Marmite’ and took the numbers regarding what the people sampled thought of it. And ate some toast.


Figure 1. Numbers of people with certain opinions regarding Marmite.

Figure 1 shows the number of people who reported that they loved, liked, felt neutral about, didn’t like or hated Marmite. The actual YouGov website actually shows picture representations of heart, smiley jaundice face, straight-mouth jaundice face, sad jaundice face and angry rosacea face that I interpreted to mean the aforementioned categories. I’m good at emoticons; sideways punctuation smiley face.

You can see that the two tallest bars are for Love It (3,289 people) and Hate It (2,235 people), followed by Like It (1,870 people), Neutral (1,067 people) and Don’t Like It (909 people). However, these aren’t necessarily the groups we’re interested in. The claim is that people either love or hate Marmite. Figure 2 shows the number of people of love or hate Marmite (Love It plus Hate It) and the number of people who don’t feel that strongly about it (Like It plus Neutral plus Don’t Like It). Of the two populations, Love It or Hate It (5,524 people) is larger than Don’t Feel That Strongly (3,846 people). This is perhaps shown more intuitively in Figure 3, where it is depicted that compared with people who don’t feel that strongly about Marmite, 17.9% more people love or hate it.


Figure 2. Numbers of people who love or hate Marmite and who don’t feel that strongly.

The presence of a group of people that don’t feel that strongly about Marmite would seem to contradict the idea that there are only two populations with respect to Marmite desire. However, it could be argued that we are really examining the effect of Marmite on Marmite apathy. Does Marmite have an effect on whether you love or hate it or don’t feel that strongly about it? What is the probability of this many people loving or hating Marmite if Marmite doesn’t make you love or hate it?


Figure 3. Proportions of people who love or hate Marmite and who don’t feel that strongly.

As this was a single population (people who give their opinions to YouGov UK) and we are looking at two possible categories within that population (Love It or Hate It and Don’t Feel That Strongly About It), I used a binomial test to determine the probability that there was on effect of Marmite on Marmite emotiveness. This demonstrated that the chances of this many people loving or hating Marmite if Marmite doesn’t make you love or hate it was at least 1 in 100,000,000 (P<0.00000001). Depending on your threshold for such things, this would seem to be reasonable argument that Marmite has a tendency to make people feel strongly about it.

There are some potential problems with this reasoning. Firstly, the analysis could be wrong. I’m far from an expert in statistics, and it’s entirely possible that I performed the wrong tests or interpreted the results incorrectly. While eating toast.

Secondly, these data only covers people who provide information to YouGov UK. While YouGov UK would certainly claim that they are representative of the whole population, we can’t know this for sure. The same YouGov UK page claims that being a Marmite customer correlates with having gardening as a hobby and being a customer of Waitrose, and I can count on the finger of know hands the number of times I’ve seen someone pruning the roses, while eating a Marmite sandwich and some Waitrose pickled quail eggs. This is a real product, although I think it’s cruel to pickle quails. Although, that’s not really the issue. Ultimately, there might be something different about the people who report to YouGov (such as a tendency to feel strongly about yeast-derived devil’s treacle) compared with the general population, and we can’t know that just from these results. Basically we’re saying that these results may be influenced by self-selection bias.


Well, what would you have put a picture of? Photo licensed under the Creative Commons Attribution-Share Alike 3.0 Unported license.

When people are in groups, their opinions and behaviour have a tendency to be more extreme than when they are acting as individuals. This is known in psychology as group polarisation. For example, if you have racist and sexist attitudes and join a group with racist and sexist attitudes, your racist and sexist attitudes will worsen; the group influence will trump your own lesser tendencies. Ahem. This process has also been seen to occur through social media, even though people aren’t physically interacting as groups. Observed over time on Twitter, discussion regarding political issues with like-minded individuals becomes more homogeneous and more extreme. In this instance, the hypothesis is that people identify with others who have a similar opinion to theirs regarding Marmite, and over time polarise that existing opinion until they state that they love or hate it. In reality, the truth is closer to a more moderate Marmite approval or disapproval. However, the online poll doesn’t involve group discussion and polls are completed anonymously, so even if people are basing part of their social identity on how much they enjoy a salty brown loaf goo, group polarisation seems unlikely.

Of relevance here may be a type of response bias called, ‘extreme responding’. This is a tendency for people to select the most extreme responses available to them and usually depends on the wording of the question, but has been linked to age (younger = more extreme), educational level (lower = more extreme) and cognitive ability (lower = more extreme). We don’t know how the poll was worded or the composition of the poll responders, so speculation as to the extent of extreme responding is fairly pointless even though it DEFINITELY HAPPENED!

Alternatively, the well-known advertising for Marmite may have introduced another kind of response bias called ‘demand characteristics’. Here, participants in an experiment or survey change their response because they are in an experiment or survey. This is assumed to be an attempt to comply with what they believe the aims of the experiment to be. Respondents asked about Marmite may be more likely to give an extreme response based on the advertised ‘consensus’ that people either love or hate Marmite. And so the opinion spreads like a pun-based analogy.

Finally, it could actually be the case that Marmite has such a distinct flavour that people really are more likely to have an extreme response than an ambivalent one. Although at this stage you may have stopped caring. I prefer jam anyway.

Does Sean Bean Always Die at the End?

The Alpha Sean Bean, shown here to be still alive. The Alpha Sean Bean, shown here to be still alive.
“Sean Bean TIFF 2015” by NASA/Bill Ingalls. Licensed under Public Domain via Wikimedia Commons .

There’s a quote from a character in The Lord of the Rings: Fellowship of the Ring, and J.R.R. Tolkein’s character from some book or other, that has been doing the rounds as an internet meme for quite some time: “War makes corpses of us all.”  Of course you all know it, it’s ridiculously famous, after all, one does not simply forget a Faramir quote. Much better than Boromir. In Sean Bean’s case however, the quote might as well be “appearing in a role in television or film makes a corpse of me, Sean Bean.” Sean Bean is well known for dying in films. So much so, that there exists a campaign specifically against the further onscreen killing of Sean Bean. At least, I think it still exists. It might have died.

Basically it is a fairly common assumption that if Sean Bean is in something, he will most likely not make it to the end. However, everyone knows what happens when you assume; you make a prick of yourself. Is it actually true that Sean Bean always dies? In psychology, confirmation bias describes the tendency for people to better recall information that confirms their existing beliefs than information that would refute them. The frequency illusion is where something (it can be an event or just an object) which has recently been brought to a person’s attention suddenly seems to occur or appear with greater frequency than it did before it had been noticed. This is also known as the Baader-Meinhof Phenomenon and once you know about it, you’ll start seeing it everywhere. So it is possible that the appearance of Sean Bean’s repeated celluloid mortality is a function of some common cognitive biases rather than him actually ending more times than a Sunday furniture sale. The following information that was collected to test this may contain spoilers for Sean Bean projects. Unless you believe the appearance of Sean Bean in a cast list is in itself a spoiler.

Using some sort of internet search engine (if you want to find a similar one, you can look it up on Google) all of Sean Bean’s roles in film and television were listed to create a population of Sean Beans. From here forward, the collective noun for Sean Beans used will be “population” rather than the perhaps more common “can” or “cemetery.” Sean Bean’s roles in theatre or performing voiceover in video games were not included due to a combination of being too difficult to include, laziness and the words “Sean Bean” starting to lose all meaning. The actual actor Sean Bean (the Alpha Sean) was also included, as while technically it is an ongoing role, we do know with reasonable certainly that Sean Bean will die at the end of it. The Alpha Sean was not included in any cause of death calculations in case I end up as a suspect in a future murder investigation. Jupiter Ascending was not included for obvious reasons.

The number of times Sean Bean was dead at the end of a film/TV show and the number of times Sean Bean was alive at the end of a film/TV show were counted and used to calculate the incidence of death for the total population of Sean Beans. The incidence rate is the number of new cases of a disorder or death within a population over a specified period of time. This is commonly express in terms of per 100,000 persons per year. In terms of deaths, this in some ways can be seen as equivalent to the Mortality Rate. Some basic demographics, causes of deaths and intentionality of deaths were also calculated.

The demographics for the population of Sean Beans are shown in Table 1.

Table 1. Sean Bean Demographics

Characteristic Sean Bean Numbers
N 75
Mean (SD) age, years 6,0810,851.05 (523,114,369.60)
Species, n (%)
Actor 1 (1.33)
Human 71 (94.67)
Lion 1 (1.33)
Portrait 1 (1.33)
God 1 (1.33)
Alive, n (%) 45 (60.00)
Dead, n (%) 30 (40.00)

The incidence of Sean Bean deaths across the total existence so far of Sean Beans (6000 BCE to 2072) is 4.85 per 100,000 person per year. The causes of Sean Bean death and intentionality of Sean Bean death are shown in figures 1 and 2, respectively. The most common cause of death was being shot by a gun. The best cause of death was fall from cliff due to a herd of cows. Most Sean Bean deaths were intentional (as a result of homicide) compared with accidental and orcicide.

Figure 1

Figure 1. Cause of Sean Bean death.

Figure 2

Figure 2. Intentionality of Sean Bean death.

The aim of all this Beanian death numbering was to determine if there was any truth to the common belief that Sean Bean always dies at the end. Examination of a fairly complete population of Sean Beans shows that this is not the case, with 60% of Sean Beans managing to survive the time it takes for many film and TV directors to tell a story. If you are a Sean Bean though, it seems you are most likely to die by being shot by a human. There may be some money to be made in a line of Sean Bean-specific bullet-proof vests.

So why is the belief that Sean Bean always shuffles off the mortal coil at the end so common? The application of confirmation bias to this has already been discussed, but for that particular bias to take effect, there must be an existing belief to confirm. The earliest manifestation of Sean Bean’s tendency for premature televisual corpse shenanigans that I could be found was approximately around his fourth appearance. However, at a preliminary glance, Sean Beans don’t seem to kick the bucket particularly often early on in the ascendance of Sean Beans to make any reputational impact.

If we divide the appearance of Sean Beans into tertiles (an ordered distribution divided into three parts, each containing a third of the population, not an aquatic reptile with a shell) and look at the proportion of deaths as time progresses, we get something that looks like figure 3.

Figure 3

Figure 3. Proportion of Sean Bean deaths by Sean Bean time tertile.

We can see that if 3 is the most recent tertile and 1 is the furthest in the past, then the Sean Bean death rate appears to be greatest in the middle of the population’s progression through time. In psychology, the serial position effect describes the tendency for people to recall items earlier (the primacy effect) and later (the recency effect) in a list the best, with items in the middle being recalled the least. This would not explain the Sean Bean always dies reputation, as in such a model we would expect more deaths in the first and last tertile. Besides, one explanation for the serial position effect is that earlier items are stored more effectively in long term memory than the other items, while more recent items are still present in working memory and are thus easily available for recall. This would only apply to these data if people experienced Sean Bean necrosis as a list in front of them, which most people (besides me) don’t. Even if the data matched a serial positioning explanation, it would be a stretch (i.e. wrong) to use it to explain the Sean Bean deceased at the finale reputation phenomenon.

Rise of the Nicole Kidmen would be a good episode of Doctor Who. Rise of the Nicole Kidmen would be a good episode of Doctor Who.

Characters don’t become instantly well known in popular culture. It takes time for a reputation to build and saturate society. In this respect, perhaps we can consider the middle tertile to be more akin to the starting point for a reputation i.e. Sean Beans will be more well known, with more opinions being formed about them. The Sean Bean death rate here is 52%, meaning that during this period Sean Beans were slightly more likely than not to die at the end. This may be enough to start the rumour of Sean Beans’ non-existence by the credits and establish a source for confirmation bias.

Characters don’t exist in isolation. They usually exist in a complex ecosystem of other populations. The Sean Bean population exists alongside the population of Bruce Willises (Willi?) and the population of Nicole Kidmans (Kidmen?) among others. Important data to consider would therefore be how often Sean Beans die in comparison to other populations. If the comparative death rate of Sean Beans is noticeably higher than that of other comparable populations, then this may explain the Sean Bean clog-popping conundrum. Future “research” should focus on this (I can’t be bothered right now).

It was suggested to me by KTBUG (kgwright73) that the popularity of the mode of presentation of Sean Bean would have an impact on the perception of his tendency for pushing up the daisies. It seems feasible Sean Beans die in more popular things and live in less popular things then the public perception would be that of a gentleman prone to leaving his life behind. To this end (where available) I took an average of lifetime box office takings for films where Sean Bean died and films where Sean Bean lived (figure 4).

Figure 4

Figure 4. Average lifetime box office takings by Sean Bean survival.

Figure 4 shows that films where Sean Bean shook hands with the Grim Reaper on average took more at the box office than films where Sean Bean continued respiring. If we use this as a crude measure of popularity (and it is very crude, subject to bias from missing TV shows and films where I simply couldn’t get the info) and impact on cultural awareness, then films where Sean Bean becomes an ex Sean Bean seem to have made a larger cultural impact. This could certainly be at least one source of the idea that Sean Bean always dies.

Please note, I am in no way suggesting that Sean Bean dying in it makes a film popular. As the old saying goes, “Sean Bean’s death correlation, does not prove film popularity causation.” You all know it.

In conclusion it would seem that Sean Bean’s reputation for always dying at the end is somewhat over exaggerated, with a death rate of approximately 40%. Sean Beans are most likely to die from being shot intentionally by a human or from being in the middle of their career trajectory. The Sean Bean Ex-Parrot Meme may be best explained by a high death rate at a time when Sean Beans were likely to be reaching their maximum prevalence in the public eye and by films which feature a Sean Bean death having made a larger cultural impact than films that feature a living Sean Bean at the end. These perceptions feed into confirmation bias. And then Sean Bean died.