Any historian dealing with sources in a foreign language will tell you a different story, but it might be instructive to consider a case in which a historian's interpretation relies on not only a correct translation, but also on an understanding of the language in question. Idioms, proverbs, and other forms of expressions can create serious misunderstandings in the researcher who is not familiar enough with the original language of the primary source.

In 2008, Krigets skönhet och sorg (later released in English as The Beauty and the Sorrow: An Intimate History of the First World War, 2011) by famed Swedish historian Peter Englund was released. Covering the First World War in 212 chapters through the eyes of individuals, it is a work heavily based on memoirs, letters, and other personal primary sources from participants in the time period. One of those individuals that Englund selected for inclusion in his book is the Venezuelan adventurer and soldier, Rafael de Nogales Méndez.

Nogales Méndez is an interesting character to study in this context. His intention was to originally join the Belgian army (since it was the invasion of Belgium that created his desire to participate in the conflict), but no army except the army of the Ottoman Empire accepted him. Serving in the Caucasus front, he becomes an eyewitness to the Armenian genocide. When a rumor reaches him that he's going to be murdered presumably on the basis of his Christian faith, Nogales Méndez rides to Aleppo to request that he be mustered out of the Ottoman ranks. His request is ultimately denied and he participates in the last campaign of the Ottoman Empire in Palestine. Nogales Méndez becomes an eyewitness to the scarcity of food in the region, epidemics of typhus, and logistical problems in the ottoman army. Furthermore, he is also a witness to public executions of Arabian deserters, men who had been forced to join the Ottoman army (unlike Nogales Méndez). At the first and second battle of Gaza, Nogales fights as part of the Ottoman cavalry.

It is here that we reach Englund's analysis. According to Englund's analysis, Nogales Méndez becomes increasingly disillusioned by the anonymity of war. Englund specifically cites a specific sentence in Méndez memoirs in which he speaks of preparing to contribute his "grain of sand". Englund focuses on the word "grain of sand" ("sandkorn" in the original Swedish) as a sign that Nogales Méndez has found his place in the anonymity of industrial warfare where the individual can not have an impact on the big picture.

However, if we look closer at the original primary source, we get a different picture. According to Englund's bibliography, he has used the English translation of Nogales Méndez memoirs, Four Years beneath the Crescent, published in 2003. The assumption that I have is that Englund does not understand the Spanish context through the English translation. In the original memoir, Cuatro años bajo la media luna, published in 1924, Nogales Méndez uses the phrase as part of a saying: "para ir a contribuir también con mi granito de arena". Reading it in Spanish, it becomes clear that he's speaking less of disillusion and more of teamwork, of pulling his own weight, doing his fair share of work. Nogales Mèndez, in this reading, is determined to be part of the larger picture, to do his duty and to contribute to the defense of Palestine. Rafael de Nogales Méndez is proud, not disillusioned.

This specific case might be considered pedantic by some, but it completely changes Englund's interpretation of Nogales Méndez that has been popularized through his book. It shows how important it can be, in some cases, to take a closer look at a primary source in its original language, in its original context, and to bring an understanding of the language in reading it. Relying exclusively on a translation can be a risk. If you do, you are likely going to have to justify using it in your research (like all good, transparent research does!).

Medievalist here. I specialize in the Angevin Empire and historical writing in the 12th Century, so most of my sources are either in Latin or French.

When I am doing serious research, I will typically favour copies of source material that have the original language, with a facing translation. These can be prohibitively expensive, and since I am not currently in a grad program (between my Masters and PhD), I am slowly building a little collection. This is important in my field because a lot of translations in Medieval Studies were done in the 70's and 80's, when translation standards were a little different and more 'artistic license' was tolerated. A language teacher during my MLitt told me that, at least in the UK, translations are not counted as 'academic output' by UK evaluations, so new editions of translations are not being produced. She actually encouraged me to possibly help supplement my income, should I be teaching in the UK, by doing translations.

Moreover, a translation can often be coloured by the interpretation of the original translator, which can muddle the meaning of the original author. The best example I have of this is descriptions of the reign of William II of England. Some translations play up the interpretation that William Rufus may have been homosexual (which I do not disagree with), but many of the actual words are deliberately mistranslated to be more explicit. Unfortunately I don't own a Latin-facing edition of the one I remember being most egregious, but it described the King and his intimates as walking with "a mincing gait" and wearing their hair in effeminate fashions. However, the original Latin is much more subtle about it. Particularly the word "mincing" was an odd choice of translation, though I can't remember what the Latin word was.

It's best practice in Medieval history to use the original source material as much as possible. When I research, I will generally do a quick read through of just the English translation, notating passages of relevance to my given topic. Then I will do a more in-depth reading of those passages, comparing the English to the Latin. I do not read Latin particularly fluently, so a lot of reference material is required, and often I'm only looking at individual phrases. The extent to which I do this depends a lot on the quality of the translation. It gets worse if I'm dealing with a text that was originally in verse and has been translated into prose in English. These are often deeply problematic. There is a text I'm working through on my own in Medieval French prose that was translated in 1976 that was heavily embellished by the translator, but needs to be redone. I'm not doing a translation, by any means, but trying to remove the embellishments and additions in order to get back to the original quality of the text.

There are also a lot of cases where there is not a readily-available version with the original text with and English-facing translation. I actually own three separate copies of Marie de France. One that is an academic English translation of her work (with one or two poems actually having a French-facing text); second is a laymans English copy for everyday use or pleasure reading; third is the original Old French text with facing modern French translation.

Tangentially related: With Medieval texts in particular, there may be several versions of the same work. As things were copied by hand, the content and quality of an original text may depend on the skill of the scribe, and when it was written. For example: Henry of Huntingdon was incredibly critical of King Henry I after his death, but revised his entire manuscript with Henry I's grandson, Henry II, came to power. A good translation should have both the original text, and the edited version, or should notate where major copies diverge and discuss why this divergence happened.

Sorry for the long post!