People’s history and distribution building
How hungry were Gulag prisoners? To answer this question, historians need to do two things: build a distribution of hunger in the Gulag and summarize this information for their readers. Good historians use various sources to achieve this goal. In this post, I’ll focus on only one of these sources—memoirs—as the basis for distribution building.
To simplify things even further, we will think of a memoir as capturing only a single person’s experience. Given this simplification, the historian’s task is theoretically easy: as each person’s experience is a single data point, collect as many memoirs as you can. That is your distribution, which you can then summarize for your readers.
To illustrate some of my arguments throughout this post, I’ll use examples from Anne Applebaum’s excellent book Gulag: A History.
The true distribution of hunger
Imagine that the distribution of hunger shown below is the true distribution, with each black dot corresponding to a hypothetical Gulag prisoner. In this example, some starved to death, most were very hungry, some were moderately hungry, and a small minority were well-fed.
Historians have no way of knowing the true distribution because they cannot directly observe it. They need to somehow construct it—a task fraught with difficulties.
Ways distribution building can go wrong
Distribution building can go wrong in two ways. The first one is that memoirists are not a representative sample of prisoners. In other words, certain parts of the distribution are over- or undersampled. The second one is that, even though memoirists are like other prisoners, what’s written in their memoirs is not their true experience.
Self-selection: the memoir writer is atypical
An extreme case of undersampling is when parts of a distribution are fully missing. For instance, imagine those at the extreme hunger end of the distribution. These people wrote no memoirs because they had starved to death in the camps.
Moving beyond the extreme case, not everyone was equally likely to write a memoir. Many memoir writers, such as Aleksandr Solzhenitsyn, Varlam Shalamov, and Evgeniya Ginzburg, had intellectual and literary inclinations before they wound up in the Gulag. Far fewer in numbers were memoirs by criminals or other demographic groups such as illiterate peasants. This wouldn’t be a problem if all demographic groups experienced similar levels of hunger. But they probably didn’t.
Consider intellectuals. It could be the case that they had less food than other prisoners. The hypothesis would go like this: In many camps, criminal prisoners held substantial power, which included control over access to better jobs and better food. Intellectuals were political prisoners and were hated by the criminal prisoners. As a result, the criminals prevented intellectuals from obtaining better jobs and sourcing better food.
Alternatively, we could have intellectuals suffer less from hunger than other prisoners. It could be the case that a trait that had turned these people into intellectuals in the first place (e.g., creative thinking, the ability to navigate the social landscape) also helped them in the Gulag. These traits allowed them to get better jobs and more food. In line with this hypothesis, Applebaum mentions that virtually all the widely read memoir writers were trusties who held good jobs at some point during their imprisonment. In other words, memoirists were probably overrepresented in the more fortunate half of the distribution.
The truth and nothing but …
Imagine we managed to get hold of memoirs by a representative sample of prisoners. Having overcome one obstacle, we now face another: only the memoir writers know how much they suffered from hunger. What ends up on the pages of the memoir might not have been their true experience—if their experience ends up in the memoir in the first place.
Mere omission
One possibility is that the person simply didn’t write about their experience with food and hunger. A memoir cannot cover all aspects of life. Maybe the food situation wasn’t remarkable, so the writer didn’t feel the need to mention it. In this specific example, I don’t think omission was likely because food availability was top of mind for pretty much everybody in the Gulag.
In other cases, however, writers might have avoided certain topics because they considered them taboo. Applebaum mentions male homosexual behavior and suicide as examples of topics that few memoirists wrote about.
Faulty memory
Someone could in theory misremember the level of food scarcity. Some want to actively forget what happened to them. Others recount their stories when given the chance. Although through different mechanisms, both can lead to faulty memories. As with mere omission, I don’t think this was one of the important ways in which memoirs might have misrepresented the truth.
Social desirability and its relatives
Social desirability bias and other strategic incentives might prompt writers to misrepresent their experience. Hungry people go to extreme lengths to obtain food. They might steal the dying person’s last piece of bread, commit other immoral things, or engage in activities they find shameful after the fact. After returning to civilian life, their survival strategies might seem unexplainable even to them. Many certainly don’t feel a strong desire to explain some of these events to their friends, let alone strangers. So, they sanitize the truth or outright lie. Even without morally questionable acts, someone might exaggerate food scarcity to gain sympathy or to conceal their relatively privileged position in the camps.
In a large enough group, there are inevitably true believers as well. These are people who identify with the ideology of the regime that sent them to the Gulag. Even if they had almost starved to death in the camps, they felt the need to downplay food scarcity in their later writings to protect the image of the regime.
Putting it all together
These are just some of the factors that make the historian’s job difficult. The true distribution of hunger is unobserved. If social desirability and other factors introduce a systematic error (e.g., every memoir either exaggerates or downplays food scarcity), having more and more memoirs will not move us closer to the true distribution. We will just be more confident in our biased constructed distribution.
It’s difficult to have an intuition about how all the different factors might bias our constructed distribution. Many good historians have different intuitions: they use different heuristics to correct for over- and undersampling, memory issues, strategic incentives, and other potential problems. This means their reconstructed distributions can differ even when based on the same materials.
A memoir as a distribution
Throughout this post, I assumed that a memoir gives us a single data point. This assumption is unrealistic. Memoirs often narrate not only the life of the memoirist but also the lives of other people the memoirist met in the camps or heard about through connections. For example, for his Gulag Archipelago, Solzhenitsyn interviewed 226 fellow prisoners about their experiences. So, a memoirist also builds a distribution. Consequently, the historian’s job could be reframed as aggregating distributions rather than single data points. Whether aggregating single data points or distributions, the issues that potentially plague data collection remain the same.
Conclusion
As readers, it’s useful to think explicitly about how a distribution presented by a historian might differ from the true (unobserved) distribution. Good historians are aware of all the ways in which data collection can go wrong. They are transparent about missing parts of a distribution, under- and oversampling some groups, and other potential issues. They try to corroborate questionable memoir content via other sources.
We might not be able to gauge the direction and exact magnitude of the bias in the constructed distribution. This would often require detailed domain-specific knowledge we don’t have. What we can do, however, is judge the credibility of the historian, discarding claims by those who, through incompetence or another reason, fail to consider relevant ways in which data collection and interpretation can go wrong.