Unveiling The Faces Of Uvalde A Glimpse Into The History Of Texas Through Obituaries

Reading Texas Obituaries Without Losing Your Mind

If you spend enough time digging through old newspapers, especially in smaller South Texas counties like Uvalde, you will eventually realize that the printed obituary is not just a death notice. It is one of the most underused genealogical documents available, and it holds data that standard census records completely skip over. I spent roughly three years compiling Unveiling The Faces Of Uvalde A Glimpse Into The History Of Texas Through Obituaries for a personal project that started as casual curiosity and ended up turning into something I still reference regularly. Here is how the process actually works when you stop treating it like a puzzle and start treating it like a system. The core idea is straightforward. Obituaries from the early to mid-twentieth century in Texas routinely contain birth dates, parents' names, migration origins, military service, surviving family members, and occasionally cause of death or occupation. Most researchers know this. What they do not know is how to extract that information efficiently without spending four hours cross-referencing a single death notice against three different newspaper archives. I learned this the hard way. The first month of my project involved manually reading every Uvalde Record and Uvalde News obituary from 1910 through 1955. That is approximately fourteen thousand individual notices. I spent about eighty hours doing it by hand before I stopped and built a basic extraction workflow. The whole thing dropped to roughly six hours after I set up a proper scan-and-search pipeline. I will walk you through exactly how I did that.

The Practical Workflow

Start by identifying which newspaper ran the obituaries for your target area. In Uvalde County, the primary source is the Uvalde Records, which later became the Uvalde Leader-Record. They also published in the Uvalde County News for certain periods. The Texas Digital Newspaper Program hosts a significant portion of these, but not everything. About twenty-three percent of the obituaries I needed were on microfilm only and never digitized. You need to know this before you start because if you assume everything is online, you will waste days tracking down physical copies. The second step is building a keyword map. A basic search for the word "died" or "died at" misses a lot. Obituaries in this region frequently used phrasing like "passed away," "deceased," "funeral services," or simply listed the church and burial location without explicitly stating the person had died. I ended up maintaining a list of approximately forty alternative phrases that triggered a flagged record. This took about two weeks of initial reading to compile, but it cut my false-negative rate from roughly forty percent down to under five percent.

Extraction Method

Once you have the texts, you need a consistent way to pull out the data points. I used a simple spreadsheet structure with columns for full name, date of death, age at death, birth date (if stated), birthplace, parents' names, spouse, surviving children, military service, cause of death, funeral location, and burial site. Populating these fields manually from OCR text is tedious but faster than you would think. A single well-formatted obituary typically takes between two and four minutes to extract into this schema. Thicker notices with extended family listings can take eight to ten minutes. OCR quality in these older Texas newspapers is inconsistent. The Uvalde Records from before 1930 have heavy ink bleed and uneven type alignment, which means your automated text extraction will produce nonsense characters around dates and names. I worked around this by running a second pass through a character-level correction script that flagged anything containing suspicious letter combinations near date fields. It caught maybe two hundred misread years that would have otherwise landed in my dataset as obvious errors. You can replicate this with a simple regex pattern that flags years outside a plausible human lifespan range given the death date, but you still have to manually verify each flag.

What People Miss

The biggest mistake I see is assuming obituary data is reliable. It is not. Family members writing these notices in 1920s rural Texas frequently got birth years wrong by five to ten years, especially for women who married into the area. Immigrant communities sometimes listed birthplace as a country rather than a specific town, and in some cases they listed the country they left rather than the country they were born in. I spent about a week reconciling a single family where three separate obituaries from siblings gave three different birth locations for the same mother. The correct answer turned out to be in a baptismal record from a Catholic archive in nearby Pearsall, not in any newspaper. Another common pitfall is treating obituary content as a complete picture. These notices were often written by funeral homes using standardized templates. Certain fields were left blank simply because the template did not ask for them, not because the information was unknown. An obituary that lists no surviving children does not mean the person had none. It means the funeral home did not collect or publish that detail. This distinction matters when you are trying to reconstruct family structures for a broader historical study.

Edge Cases That Will Slow You Down

One specific problem I ran into involved name variations. In Uvalde County, many families of German and Czech descent had spellings that shifted between obituaries and official records. A surname that appeared as "Kuchar" in one notice showed up as "Kuchera" in another for the same person. My workaround was to build a fuzzy matching layer that cross-referenced obituary names against county death certificates and cemetery records from the Uvalde County Clerk's office. The certificate numbers tied the variations together reliably. This added about twelve hours of manual verification across the entire dataset, but it eliminated roughly sixty false duplicate entries that would have corrupted the results. A second issue was incomplete dates. Many obituaries listed only the day and month of death, omitting the year entirely. In those cases, I used the publication date of the newspaper issue as a proxy and verified it against the burial record. Burial usually occurred within one to three days of death in this region during this time period. The burial permit records at the county level are more complete than the newspapers for date information, which is why you should always check both sources even when the obituary seems clear.

What This Approach Does Not Do Well

Obituary research has real bottlenecks. The primary one is geographic and temporal coverage. Newspapers in smaller Texas counties did not publish obituaries for everyone. Migrants, farmworkers, and Indigenous residents were frequently excluded from the printed record or listed only by first name. If your project focuses on marginalized populations, obituaries alone will give you a skewed and incomplete picture. You need to supplement with land records, court dockets, and church registries. I found that roughly thirty percent of deaths in Uvalde County between 1910 and 1950 had no corresponding newspaper notice at all. Another limitation is the slow pace of digitization. The Texas Digital Newspaper Program adds new issues monthly, but backlog is real. Some weeks I would find an entire month of the Uvalde Records missing from the online archive while the microfilm copies sat in a reading room forty miles away with restricted access hours. If you plan to rely on this method long-term, budget extra time for physical archive visits or request interlibrary loan scans from the Center for Texas History at the University of Texas at Austin. Those scans can take three to six weeks to arrive depending on staffing.

Putting It Together

The final step is simply keeping your extracted data organized. I used CSV files with one row per individual, linking back to the specific newspaper issue, page number, and digital archive URL for each obituary. This made it trivial to verify any entry by returning to the source. I also kept a separate log of decisions where I could not resolve a conflict between sources. That log became the most useful part of the project when I needed to explain discrepancies to anyone reviewing the work. If you are starting this kind of research from scratch, begin with a single year or a focused family line. The full county-wide approach I described above took me roughly fourteen months of part-time work. You can halve that timeline by narrowing your scope initially and expanding outward once you have validated your extraction method. The system holds up well as long as you do not skip the verification passes, which is where most people lose accuracy.