This article is based on a telephone conversation between Sally Miles, Ph.D., CCC-SLP and Karen Andriacchi M.S., CCC-SLP. The conversation was a follow-up to the language sample transcription [by SALT Services] and the analysis [by Sally using SALT software] of the Brendan Dassey interrogations made famous by the Netflix documentary, “Making A Murderer”.
The first half of the interview (today’s post) focuses on how SALT can be used by researchers for non-standard uses. But we think that clinicians will find many of these techniques useful as well. When we can, we have pointed out how these strategies might be applicable to the wider community of SLP practitioners.
The second half of the interview (coming in the near future) will be a bit different for us. It focuses on the role of SLPs as advocates for our clients. That post won’t be very “SALT-y,” but we hope you’ll tune back in anyway. As Sally demonstrates, SLPs have the potential to shine light into some very dim corners of the American criminal justice system.
The interview has been edited for length and clarity.
I would like to hear the story of how you met attorney Michele La Vigne and decided to get involved in the Brendan Dassey case.
We knew each other just a little because we live in the same neighborhood. At a neighborhood party, Michele mentioned that she had recently given a talk about language disorders and the law: that one of the problems for lawyers is that people with language disorders have poor narrative skills, so it’s difficult for the lawyers to get information about the case from their clients in order to develop a defense.
I had no idea lawyers knew about language disorders! I told her I was an SLP and that I had done my dissertation on narrative development in children and had a long-standing interest in narratives. We started trading research and discussion. I thought it would be interesting to work together but couldn’t see how to make that happen unless there was a case that I could help with. But Michele is a professor, not a practicing [trial] lawyer, so there was no case. Until Making a Murderer came out.
Michele watched it but I wasn’t interested. But one day she asked me to look at a section of clips showing the interview of Brendan Dassey, who was known to have cognitive and language disabilities. She wanted my impressions as an SLP. I watched it and was blown away by what the police were doing: feeding Brendan the information they wanted him to confess to.
The interview method the police used was entirely inappropriate for someone, like Brendan, who has language disabilities.
The police officers who interviewed Brendan had been trained in what’s called “The Reid Technique,” a method of interviewing that one of Brendan’s current lawyers describes as a way “to dominate and control the narrative.” I think the police officers were misled by their training into thinking that they had this fabulously powerful interview strategy. (It is good at getting false confessions!)
There is a need for more – or different – training for these people in law enforcement.
Absolutely. As we proceeded through this project, I kept asking Michele, ‘are there any guidelines for interviewing kids?’ And the real shame of it is that there is training out there – my literature search revealed that psychologists have developed protocols for interviewing children that have been very carefully constructed to avoid contamination.
But not all officers get that type of training. The mission here appeared to be to get a confession or somebody locked up. Is that right?
It certainly seems that that was their goal, rather than getting information. That’s an important point we wanted to make in our study: an interview is a very specific type of communicative interaction that is supposed to be neutral and information-seeking. It is a skilled professional interaction that requires training and periodic review. But the way the officers conducted the interview it was accusatory and, instead of seeking information, they planted it.
They fed him the language?
Yes. I think that when the interview started, Brendan was considered a witness, not a suspect. But then, as the interview proceeded with the strategies that they used (and I use the word strategy extremely generously), they got him to confess to things in an illegitimate way. So he morphed from a witness into a suspect. And then at the end of the interview they told him he was going to be arrested.
So, Michele asked me if I would watch a segment of one of the episodes. What was clear to me immediately was that Brendan wasn’t talking much – the police were doing most of the talking. And all of the officers’ questions contained huge quantities of content rather than being open ended and allowing Brenden to tell the story. But, you know, I just didn’t want to draw conclusions based only on watching the film. (It turned out I was right about quite a bit, for example SALT data showed that Brendan contributed only 28% of words spoken.)
Did you transcribe and analyze the recorded interrogations using SALT? Or did you just want to see it all on paper?
I just wanted to see it on paper. I didn’t have SALT software at that time. I transcribed in Word, just to have it in front of me so I could look at all the questions and responses. Then I could draw conclusions from that, rather than from just impressions.
After transcribing and doing an analysis, I told Michele that at least in those five minutes of the interview, it was conducted inappropriately making the results illegitimate. What I didn’t know was: was this excerpt representative of the entire interview? Or did the documentary just cherry pick things to support their idea that Steven and Brendan were being railroaded?
I told Michele that we needed to look at the whole transcript to be able to answer those questions. I talked with her about language sample transcription and SALT and how it was different from court or police transcription where they clean things up, smoothing out the language. And of course, police transcription does not have any means to generate language measures, so without language analysis software we couldn’t get into any substantial data analysis of lengthy interviews. I think at this point we found the full interviews on YouTube, and after listening to random parts, we saw that the excerpts were representative, and we were ready to go ahead. Michele had grant money to put into it, and we were both excited to embark on this project that had no precedent in either of our professions.
Are the full video and audio of Brendan’s interviews public record?
Yes. So all Michele had to do was to call up the county and request the records. The video and audio of the police interviews with Brendan as well as Brendan’s school records were part of public records because they had been put into the record of the court proceedings during the trial. So we immediately got his language evaluation results. The evaluation had been done close to the time period of the police interviews.
Right, his three-year reevaluation?
Exactly. Brendan’s triennial evaluation was completed just a month before this terrible murder took place. The police interviews were conducted about 4 months after the eval. So there was no doubt about whether he would have improved over time or changed significantly since the eval – we had a timely snapshot of his language.
Then we had all the material we needed to start the analysis in front of us. I contacted SALT and we set up the arrangements for them to do all the transcription.
Michele had initially wanted to use my impressions of Brendan’s interrogation in her presentations to legal professionals. But once we committed to the full analysis and SALT, then it was going to be an article. I had worked with SALT as a graduate student at UW-Madison on several projects, so I was very familiar with it and how it worked. I knew how flexible it was, too: not only could we use the existing SALT codes and conventions and get the standard measures, but we could also insert our own coding.
Custom coding! So, you are one of the users who goes beyond the standard!
Oh, yeah! I wanted to mine the standard measures and other analyses built into SALT as much as possible. And that was a really good start and a great foundation. It gave us lots of good information. But it wasn’t enough. Just using the standard measures – like the number of questions for example – was not going to tell us enough. I wanted to know how often the police used high-content vs. open-ended questions, and if they allowed Brendan to tell the story in his own words. So I coded each different type of question: open-ended, wh-questions, yes/no, and multiple choice.
So you created and added some codes for those question forms for later analysis.
Yeah. And I would think that clinicians would find that useful, too. Not just for the adult’s questions but for the child’s questions, too.
You know, I love that you said that.
Because I do try to give examples when I do trainings, outside of what the discourse analysis shows – how you can go in deeper. And it doesn’t have to be difficult. How about: ‘no response’ or ‘inappropriate response’? You get that a lot with kids on the spectrum, for example: you might get an odd response. It’s showing as a response in SALT, but it’s not showing you the content of the response or providing qualitative data.
I like how yours is a good example of SALT’s flexibility and how this person’s language could be fully explored. Obviously, things pop up right away in LSA. But the digging deeper is the part that I love about LSA using SALT. The diagnostic, the assessment, has always been the best part, for me personally, about being a speech path.
Me too! In this study, that “going deeper” you mention was what gave us the most complete picture of how the “confession” was constructed: I had data that showed that the officers’ language limited Brendan’s participation, introduced predominantly their own content, and was difficult for him to process.
The whole idea that you have these outcomes and, “Oh! Look here! Let’s go deeper”, is exciting. And SALT is the best tool to allow you to do that.
Yeah, so that was the really exciting part for me. And, I was able to synthesize information that showed up on Brendan’s standardized testing with the custom coding that SALT allows. For example, he had very poor comprehension of paragraph-length spoken language – (9th percentile!). So what I was able to do with SALT was code and count paragraph-length talk by the policemen.
Of which there was a lot.
Oh, yes. 157 paragraphs in one of the interviews.
Also I put in a code when their paragraphs were followed by questions so that would be comparable to what Brendan experienced on the subtest that he did so poorly on. So that was a nice use of standardized test results aligned with real-life language production showing how difficult police discourse would be for him to comprehend. Clinicians can mine standardized tests for language areas to pursue in greater depth through language sampling in naturalistic conditions.
Another use of custom codes: We sent a draft of the article to Brendan’s current lawyers (not his initial lawyers that were shown in the documentary), and they told us that the prosecutors claimed that Brendan was capable of resisting the policemen’s suggestions. The lawyers said to us: ‘If you can find a way to show that Brendan was not good at resisting, that would be very helpful.’
I identified all the times that Brendan ‘resisted’ the police suggestions, and then all the times he changed his answers because of their suggestions. The data showed that prosecutors had just cherry-picked certain answers to “prove” his resistance. Brendan changed his answers more often, in the face of the police’s barrage of questions, than he stuck with them.
Other features of the SALT program I relied on were: Using this wonderful technology, I could be sure I wasn’t making any mistakes in counting or in remembering the large number of language behaviors I examined. Furthermore, if I wanted examples, I could just do a search for codes. This was enormously helpful for writing the article. A clinician could do the same thing: search a transcript for a certain code (or word or utterance) and give an example in a progress report.
Exactly. And I don’t know if you know that SALT 18 and 20 have a Performance Report that writes up a narrative of the outcomes. It will pull out examples such as ‘strong use of subordination,’ or an example of an error at the word level. With Brendan there were so many errors! I would have loved to have read SALT’s performance report from one of his transcripts!
I assume you started out looking at Brendan’s language production. I assume the question was: “What’s Brendan doing”? Maybe I’m wrong.
Before I saw the show, I assumed I would look mostly at Brendan’s language. But!! When I saw the episode, the police were so clearly doing things that were shaping the narrative, I knew that I was going to be looking at their language too. But I didn’t realize how much I would be looking at their language, how much focus that would get. Yeah, that was a surprise! As clinicians we have a basic understanding that a speaker’s language affects the other speaker’s language, so it makes sense to look at both speakers. Even more so when it’s a high stakes situation like this.
So this is what came to mind when I was reading your article: it’s almost like the police were trying to tax Brendan’s working memory. The officers mazed so much, including all those filled pauses. They would begin and then abandon utterances so frequently. And they didn’t maintain a topic.
I know! If they were doing it on purpose, they were really good at it, right?
The officers’ language was extraordinarily incoherent and disorganized. In the paper I gave numerous examples because the quality of the police language was almost indescribable; I felt incapable of summarizing it in a few descriptive terms. I could say: ’they were incoherent and disorganized,’ but you wouldn’t really get a feel for it unless you read some of the actual excerpts. The examples show how they would start with one topic, not even complete a sentence, and go on and bring up multiple other topics, and insert partial metaphors – and all at great length! Examples were the only way to show how disorganized and fragmented their discourse was.
Wasn’t it! And I was having a hard time as a diagnostician. My memory – visual and auditory – isn’t good enough to follow these people!
I know! And imagine if you had cognitive impairment, memory problems, and a language disorder. I felt so bad for Brendan! Sometimes I had to take long breaks from working on the transcripts because I couldn’t face the cruelty and ignorance that was playing out in them. This was one of the most emotionally challenging projects I’ve worked on.
Yet on the other hand, it was exciting to break new ground in language analysis and in my unusual collaboration with law professor Michele La Vigne. After all the hard, long work, it was deeply satisfying that Brendan’s lawyers considered the paper groundbreaking and used it in their appeal for clemency. The article has been downloaded hundreds of times, so it seems the word is getting out.
Want to read the full article? Find it here:
“Under the Hood: Brendan Dassey, Language Impairments, and Judicial Ignorance,” 82 Alb.L.Rev. 823 (2019). Univ. of Wisconsin Legal Studies Research Paper No. 1473