How Accurate is the Ellis Island Database Transcription?
– Megan Smolenyak

Last April, the estimated 100 million Americans having at least one ancestor who entered the U.S. through Ellis Island cheered when 22 million records from 1892-1924 were uploaded to the Internet for free searching at: www.ellisislandrecords.org

Finally, we could search for our immigrant ancestors at home in our bathrobes! Better yet, once we found them, we could view a digital image of the original ship's manifest in a matter of seconds. For longtime genealogists especially, this was an unimaginable dream come true.

Not surprisingly, there were a few kinks to be worked out. Unexpectedly high traffic prevented many eager roots-seekers from getting into the site when it was first launched. Some who did get in found ancestors, only to discover that they couldn't view the manifest due to missing or broken image links. And others complained that they couldn't find their ancestors no matter how hard they tried.

This last issue resulted in some grumbling about the accuracy of the transcription work behind the Ellis Island database (EIDB), and I found myself wondering how legitimate such gripes were. After all, transcribing millions of foreign-sounding names and places logged in often difficult-to-decipher handwriting on pages that were frequently faded, smeared, or otherwise damaged sounded pretty darn challenging to me. And we all know that many of our ancestors' names were somewhat distorted in the manifests themselves. Maybe so-called mis- transcriptions were actually faithful replications of misspellings in the original records. Considering that this was a volunteer effort, I wasn't about to look a gift horse in the mouth.

Still, I felt it was important to take a closer look. As I have extensive Ellis Island research to do, I wanted to know what the odds were that I still had some ancestors hidden in the crevices of the EIDB that I might need to get more creative to unearth. My early searches had turned up immigrants with some startling spellings of their names, so I was curious to see if I could determine roughly how much of the strange spelling phenomenon could be attributed to the way the name was recorded at the time of immigration and how much might have its origins in the recent transcription.

I decided to conduct a little experiment using my own surname. Smolenyak works well as a case study because it has that foreign sound of many Ellis Island immigrants and can be misspelled in countless ways. If anything, I thought that it would provide a very rigorous testing of the EIDB because it is more prone to misspelling than most names. I also conveniently had paper copies of Smolenyak arrivals obtained through traditional, microfilm research at the National Archives.

I began my experiment by taking the paper copies and transcribing the Smolenyak entries as best I could, letter by letter. Then I used the EIDB to pull up these same pages. Since there are actually no entries for "Smolenyak" in the EIDB, I did this with a little reverse- engineering, by locating someone else on the hard copy pages whose name was fairly obvious and searching for them. Once I found the page that contained this individual in the EIDB, it was only a matter of scrolling to find the Smolenyak on the same page. By clicking on the "view text version manifest" option, I could see how the Smolenyak name had been transcribed.

In all, I had fifteen of my own transcriptions to compare to the EIDB ones and was very pleased to find that ten of the fifteen matched perfectly. Of the remaining five, three of the mismatches were easily understood. Inspecting the original manifest, for instance, I could see how Smoleniak could be interpreted as Smolenick. The last two both surprised and impressed me: Szuwlyenak and C . . . oleniak! Using my surname as something of a barometer then, I concluded that about two-thirds are flawlessly transcribed and another 20 percent are very close. The last 13 percent were apparently going to be little more challenging to find.

My conclusion? I don't want to overstate my interpretation since the findings were based on a single surname, but I share it here as at least one indication. Yes, there are some names in the EIDB that are apt to be elusive unless you get creative. But given that I used a name that practically invites misspelling for my experiment, I would surmise that most names are completely or fairly accurately transcribed in perhaps 90 percent of the cases. And when you factor in that I had the advantage of being familiar with variations of my own name and Eastern European spelling quirks, as well as the luxury of not having to transcribe thousands of other entries under deadline, I can only express my admiration for the volunteers who logged countless hours building this database and still managed to be remarkably precise.

But just in case you're worried about that last ten or so percent, Stephen P. Morse's One-Step search form, mentioned previously in a pair of Michael John Neill's articles, is the just the tool you need. In a future article, I'll tell you about another experiment I did to gauge its effectiveness as compared to traditional, microfilm research and using just the EIDB. In the meantime, I'd be delighted to hear from any fellow EIDB experimenters about results they achieved with other surnames.

– Megan Smolenyak

Click here to access the world's largest online collection of historical records free for 14 days.