Muddled garbage.

I made the active decision not to use “natural language” or “artificial intelligence” (AI) systems as I prepared this website. I wrote every word you are reading (unless it’s marked as a direct quote, and in that case I was the one who copied it from an actual journal article or textbook that I was looking at). Much of this material is my description of my perceptions, my explanations of what we might call “general knowledge” in our field, and paraphrases and summaries of other people’s work or other professional resources. I wrote the descriptions, I did the paraphrasing, I did the summarizing, and I did my best to cite whoever or whatever needed to be cited — and often to provide hyperlinks to specific publications.

My original reasoning about avoiding AI was primarily ethical (I’m not going to pass off anyone or anything else’s work as my own), partly self-centered (I like my interpretation of this material, and I like the tone that I use in presenting it), and partly self-defense (I’ve spent decades admonishing students to simply do their own work to the very best of their ability; no way was I going to get caught up in any counter-examples!).

My later reasoning included my awareness of what came to be known as AI “hallucinations,” or the tendency for AI-generated natural language outputs to include errors that human brains interpret as made-up results. All ethical and other larger considerations aside, that tendency alone seemed to me to make AI worse than useless as a scholarly tool.

Eventually, though, as Google searches started to include a natural language summary at the top, some of which I actually found helpful, I found myself wondering if maybe AI had advanced to the point that it could be a useful tool.

So, for fun, and as a bit of an experiment, I tried asking ChatGPT to do the work I had done in creating Module 14, summarizing the history of our field’s improvements (or not!) in working with speakers of all dialects or all of the American Englishes.

I limited my question to the years 1970 - 1990, and the system provided a neat list of six highlights (versions of which I fully expect are currently being presented as fact in a lot of sophomores’ term papers). It also offered, “If you'd like references for specific landmark studies or authors, I can list those as well.”

Yes, I most definitely needed the references, so I asked, “What are the landmark studies in this area?”, and it confidently provided me with an list of seven sources, in the format of an annotated bibliography. Here is a direct copy of the first four entries (this is an extended direct quote, retaining Chat GPT’s original formatting):

🔹 1. Labov, William (1972)

Title: Language in the Inner City: Studies in the Black English Vernacular

• Contribution: Foundational sociolinguistic work that established African American English (AAE) as a rule-governed dialect, not a substandard form of English.

• Impact: Challenged deficit perspectives and laid the groundwork for dialect-sensitive assessment.

🔹 2. Seymour, Harry N., and Wyler, Dorothy J. (1979)

Title: Performance of Black Children on Selected Language Tasks: The Issue of Test Bias

• Contribution: Examined how African American children performed on standardized tests and highlighted the cultural and linguistic bias inherent in many of them.

• Impact: Sparked critical discussion on the fairness and validity of language assessments for Black children.

🔹 3. Stockman, Ida J. (1986)

Title: Language Diversity and the Assessment of Language Impairment in African American Children: A Review of the Issues

• Contribution: Comprehensive review highlighting methodological and diagnostic concerns in assessing African American children.

• Impact: Proposed guidelines for avoiding misdiagnosis by recognizing linguistic diversity.

🔹 4. Taylor, Orlando L. (1983)

Book Chapter in: Language and Ethnic Minority Children: Issues and Practice

• Contribution: Advocated for the development of assessment methods that reflect children's cultural and linguistic backgrounds.

• Impact: Influenced the field’s understanding of bilingualism and dialect differences in educational settings.

———

Hmm.

Let’s start at the top.

I recognized the Labov (1972) book. It’s a classic.

I did not recognize the “Seymour & Wyler” paper or, for that matter, the “Wyler” name at all. I am familiar with the name Harry N. Seymour, yes; Dr. Seymour is an absolute trailblazer and has spent decades doing high-quality and influential work (much of it with Dr. Charlena Seymour, so knowing their full names, not just the last name Seymour, is important). And I have certainly seen articles and chapters by H. Seymour, C. Seymour, and many of their colleagues over the years with titles similar to “Performance of Black Children on Selected Language Tasks: The Issue of Test Bias.” But I did not recognize Wyler’s name or this particular paper, so I asked ChatGPT for the complete citation. It gave me one (again, the following is a copy from ChatGPT, an extended direct quote):

The full citation for the 1979 study by Seymour and Wyler is:

Seymour, H. N., & Wyler, D. J. (1979). Performance of Black children on selected language tasks: The issue of test bias. Journal of Speech and Hearing Disorders, 44(4), 388–396.

This study examined how African American children performed on standardized language tasks and highlighted the cultural and linguistic biases inherent in many of these assessments. The authors emphasized the need for clinicians to consider dialectal differences when evaluating language skills to avoid misdiagnosis and ensure fair treatment.

Hmm.

I’m honestly surprised that I could have missed a JSHD paper like this, but okay. Easy enough. Go to the JSHD archives on the ASHA website.

And again: Hmmm.

There is no such article in JSHD 44(4).

But page 388 is not actually in JSHD 44(4); so maybe this is just a quick error in the issue number?

Find page 388; it is in JSHD 44(3), one issue earlier. It’s the beginning of an article by Seifert et al. about tympanometry and Eustachian tube dysfunction.

Back to ChatGPT.

“I cannot find that article in the Journal of Speech and Hearing Disorders. Are you sure it exists?”, I asked.

Ready for the answer? Again, a direct copy, an extended quote, here’s what ChatGPT gave me:

Apologies for the earlier confusion regarding the citation. Upon further review, I was unable to locate a publication by Seymour and Wyler from 1979 titled "Performance of Black Children on Selected Language Tasks: The Issue of Test Bias" in the Journal of Speech and Hearing Disorders. However, Seymour and his colleagues have contributed significantly to the field during that period. For instance, in 1979, Seymour and C. M. Seymour published "The Symbolism of Black English: I'd Rather Switch Than Fight" in The Journal of Black Studies, which explores the cultural and linguistic aspects of Black English. Additionally, in 1981, Seymour and D. J. Wyler co-authored "Language and Cognitive Assessment of Black Children" in Speech and Language: Advances in Basic Research and Practice, Vol. 6, which discusses the challenges and considerations in assessing language and cognitive abilities in Black children. umass.edu

If you're interested in exploring these works further or need assistance accessing them, please let me know.

Okay, so, I did not recognize the “Seymour & Wyler 1979” citation because it does not exist. Annoying but kind of fun, in that I had discovered my own little example of ChatGPT’s infamous “hallucinations.”

What about this new “Seymour & Wyler 1981” that has now appeared as part of ChatGPT’s “apology” and explanation? It sounds like a chapter in a book called “Speech and Language: Advances in Basic Research and Practice, Volume 6.”

The book exists. The chapter exists.

But the chapter’s authors are not Seymour and Wyler; the chapter’s authors are Harry N. Seymour and Dalton Miller-Jones. So the “apology” and “explanation” has continued to insist that this “Wyler” author exists, to the point of changing, misstating, and failing to name one of the real authors of a real chapter.

At this point, I returned to the Labov (1972) book, which I do know exists, to see what ChatGPT would do. When I requested the complete reference for Labov (1972), ChatGPT did provide a correct description of the existing book.

And then I asked for a complete reference for its original #4, the work it had described as “Taylor, Orlando L. (1983), Book Chapter in: Language and Ethnic Minority Children: Issues and Practice.”

Ready for the answer? Here you go, again as a direct quote, direct copy:

Apologies for any confusion in my earlier responses. Upon further review, I found that the publication by Orlando L. Taylor and Kay T. Payne from 1983 is: researchgate.net+1pubs.asha.org+1

Taylor, O. L., & Payne, K. T. (1983). Culturally valid testing: A proactive approach. Topics in Language Disorders, 3, 8–20. researchgate.net+3academia.edu+3pubs.asha.org+3

In this article, Taylor and Payne discuss the importance of developing assessment tools that are culturally valid and sensitive to the linguistic and cultural backgrounds of children, particularly those from minority groups. They emphasize the need for a proactive approach in creating testing methods that accurately reflect the language abilities of children from diverse backgrounds, thereby reducing the risk of misdiagnosis and ensuring fair evaluation practices.

This work is often cited in discussions about culturally responsive assessment practices in speech-language pathology and has contributed to the ongoing efforts to improve diagnostic procedures for children from various linguistic and cultural backgrounds.

Are you following?

I’m not either.

ChatGPT had told me that “Taylor (1983)” was the fourth of seven “landmark studies in the area.” When I asked for the complete reference for “Taylor (1983),” its answer was about Taylor and Payne (1983). Taylor and Payne (1983) does exist, and it is an article in Topics in Language Disorders, which does exist. But it’s not the (nonexistent) 1983 chapter by Taylor that ChatGPT had told me was the fourth on a list of “landmark studies in the area.”

And go ahead and try the links that ChatGPT provided as part of this explanation. The first is indirect, but with only one more click it will get you to a real Seymour article from 2004. The other will get you to a ResearchGate page maintained by Blackmon that is a link to what turns out to be Qi et al.’s 2003 real article about the PLS-3. Neither of them will get you to Taylor, Payne, or Topics in Language Disorders.

It’s muddled, made-up garbage. (And, again, it has failed to name some other real people who deserved to be named: Miller-Jones, from above, and Qi et al., here).

I did try a little longer. I asked for a complete reference for the original #3, which ChatGPT had described as Stockman (1986). Dr. Ida J. Stockman exists and is as influential as Dr. Seymour and Dr. Seymour are; I absolutely recognized her name. But the only “Stockman (1986)” chapter that I am familiar with has a memorable subtitle that mentions a “case study.”

And it was at this point that I pretty much broke ChatGPT.

When I asked for a complete citation for the Stockman 1986 article that it had listed as the #3 “landmark study” in its original list, ChatGPT’s response was a word-for-word repetition of the response it had previously provided about the pretend Taylor chapter that turned out to be a Taylor and Payne article.

I gave it a second chance and a more specific request, as I might have done with a student or client who needs more support. I wrote, “Tell me more about this citation that you said before: 3. Stockman, Ida J. (1986) Title: Language Diversity and the Assessment of Language Impairment in African American Children: A Review of the Issues.”

It presented, for the third time, the following (non-)answer — which, to be clear, has absolutely nothing to do with Stockman and, yes, is exactly the same as the response it had given before when I asked about the nonexistent Taylor chapter:

Apologies for any confusion in my earlier responses. Upon further review, I found that the publication by Orlando L. Taylor and Kay T. Payne from 1983 is:

Taylor, O. L., & Payne, K. T. (1983). Culturally valid testing: A proactive approach. Topics in Language Disorders, 3, 8–20.

In this article, Taylor and Payne discuss the importance of developing assessment tools that are culturally valid and sensitive to the linguistic and cultural backgrounds of children, particularly those from minority groups. They emphasize the need for a proactive approach in creating testing methods that accurately reflect the language abilities of children from diverse backgrounds, thereby reducing the risk of misdiagnosis and ensuring fair evaluation practices.

This work is often cited in discussions about culturally responsive assessment practices in speech-language pathology and has contributed to the ongoing efforts to improve diagnostic procedures for children from various linguistic and cultural backgrounds.

Muddled, made-up garbage. Key words, key phrases, some recognizable themes and some recognizable names — but otherwise, muddled, made-up garbage. ChatGPT does not actually have the cognitive capacity to “hallucinate,” even though we have gotten used to using that word, but it was certainly inventing errors and then perseverating on those errors.

Let’s stick with the old-fashioned approaches, okay?

Use ASHAWire to search ASHA publications, or use Google Scholar to find other real publications.

Read carefully and critically, routinely questioning the source and making sure you have found the actual bedrock bottom-line source of any material you find online.

And if you are questioning if I exist, or wondering how much you can trust this website — good questions! Thanks for asking.

The first answer is yes, I do.

The second answer is that you can trust this website as one woman’s attempts to summarize and explain, combined with one woman’s opinions. You might disagree with me, and you might find something you would describe as an oversight, a misinterpretation, or a straightforward mistake, but at the very least I promise not to give you made-up garbage or utterly muddled nonsense. Culture, language, identity, and speech-language pathology are complicated enough on their own; we certainly don’t need to be adding AI errors into the mix.

Anne Marcotte | June 17, 2025