Facilitated messages attributed to different people show distinctive styles, but this fails to rule out complete facilitator control

 (Cross-posted at FacilitatedCommunication.org).

I got so distracted with other pro-FC developments that have emerged since last August—Angie Kim’s Happiness Falls, Jennifer Binder-Le Pape’s purported myths about RPM/S2C, Vikram Jaswal’s latest pro-S2C paper, and some half-dozen other developments—that I’m only now getting to the second of two papers published this past year by Nicoli, Pavon, Grayson, Emerson, and Mitra, the first of which I critiqued here at the end of August.

As with Nicoli, Pavon, Grayson, Emerson, and Mitra (2023), this second paper, (Nicoli, Pavon, Grayson, Emerson, Cortelazzo, and Mitra, 2023), focuses on touch-based facilitated communication (FC) as used with individuals with cognitive and sensorimotor disabilities. (The individuals’ precise diagnoses and measured language skills remain unspecified). The authors’ other paper provided a highly quantitative task analysis of the sensorimotor skills involved in sitting in front of a keyboard and pointing to letters; this paper provides a highly quantitative stylistic analysis of a database of facilitated messages in Italian. Its goal is to explore who authored these messages—the facilitator, the facilitated person, or some combination of both.

This article purports to build on other analyses of facilitated messages that have found the texts of the facilitated individuals (or facilitatees) to differ stylistically from those of their facilitators—particularly in the more frequent appearances of unusual words and neologisms (made-up words). Such differences, the authors acknowledge, are not proof of authorship by the facilitatees (they cite Saloviita, 2018). And it’s impossible to know, they concede, what the facilitatee’s texts would look like, stylistically speaking, without the assistance of a facilitator, since none of the facilitatees can type independently. But the authors nonetheless insist that such a quantitative study can potentially answer questions about authorship.

Of course, there’s a much simpler, more definitive procedure that can answer—and has answered—that question: the controlled message-passing experiment in which the facilitator is blinded to the answers to the questions asked of the facilitatee. Dozens of such message-passing tests were conducted back in the 1990s, some 25 to 30 years before this article was written, beginning shortly after FC was introduced to the United States in the late 1980s. The answer was total facilitator control. The facilitators were the sole authors of the facilitated messages.

Despite this, FC persists, and Nicoli, Pavon, Grayson, Emerson, Cortelazzo, and Mitra (2023) open by introducing it. The authors describe FC as a technique used with people with a combination of “atypical cognitive capacity” and “sensorimotor coordination deficits” to “physically support them, for example, by touching their torso or arm as they type.” The authors add that “The neurophysiological mechanism of such motor assistance for linguistic expression is not known.”

Given what all the rigorous studies of FC have shown, namely that the literate facilitator is controlling the messages by directing and cueing (however unwittingly) the facilitatee, “not known” is bit of a stretch.

It’s not that the authors are unaware of this possibility: they acknowledge that the prevailing view among experts is that of total facilitator control via facilitator cues. But they insist that it’s just one theory about what’s happening, and they proceed to propose that the facilitative touch could serve, not to cue and control, but instead to help to reduce the “cognitive load” of the facilitated person. Their only evidence for this is their other 2023 article—the one I critiqued here.

While the authors also acknowledge the results of the message-passing tests of the 1990s, they also cite the various “alternative” tests that message-passing-averse researchers have conducted since then: measurements of eye-gaze fixations on letters (Jaswal et al., 2020); measurements of index-finger accelerations towards letters (Faure et al., 2021); and linguistic analysis of facilitated output (Bernardi & Tuzzi, 2011; Niemi & Kärnä-Lin, 2002; Tuzzi et al., 2004; Tuzzi, 2009). (For our critiques of these articles, see here and here). These various studies, the authors claim, create a more “complex picture” of what’s going on and suggest some involvement of the user (the facilitatee) in creating the message. Stylistic analysis in particular, the authors claim, allows for

a time-extended perspective on the text construction process as it operates naturally, without the insertion of experimental artifacts. The analysis of text generated over long periods of time, while the user and their partnership with facilitators evolve, provide a stronger test of individual influences than snapshot methods using arbitrary tasks in which the user possibly is not motivationally invested.

The authors do not elaborate. In particular, they don’t tell us what sort of “complex picture” or longitudinal text analysis, on one hand, or experimental artifacts, snapshot methods using arbitrary tasks, or lack of motivational investment by the facilitatee, on the other, could possibly account for the common outcome, seen in the rigorous message passing tests of the 1990s, of facilitatees typing out information to which only the facilitator had access—e.g. a picture of a key instead of a flower.

What other explanation there could possibly be—other than the simple, straightforward explanation that the facilitator is the one controlling the messages—is hard to imagine.

Despite their express reservations about this simple explanation, and despite the pro-FC track records of some of their co-authors (see Emerson et al., 2001; Grayson et al. 2012), the authors assure us that “we start from the maximally sceptical position that the facilitator is the sole [authoring] agent”. They proceed to claim that, if the facilitator is the sole author

It should not be possible, then, to find unique stylistic fingerprints associated with specific FC users.

Operating under this purportedly well-founded assumption (more on that shortly), and under their purportedly maximal skepticism, the authors set out to investigate whether children with developmental disabilities “exhibit their own stylistic signatures alongside those of their facilitators.” They look both at whether keeping the facilitator constant across different facilitatees results in one style or many; and at whether keeping the facilitatee constant across multiple facilitators results in many styles or one.

As with Jaswal et al. (2020), Faure et al. (2021), Nicoli, Pavon, Grayson, Emerson, and Mitra (2023), and Jaswal et al. (2024), it would have been orders of magnitude simpler to measure authorship through rigorous message-passing tests. And it would have been orders of magnitude simpler to measure literacy skills through direct assessments of literacy skills (can the facilitatee place the “key” label, say, on a picture of a key, and the “flower” label on the picture of a flower?). Instead, what we get is yet another fancy study: “a multi-level quantitative linguistic analysis” complete with “unsupervised machine-learning methods, particularly cluster analysis.”

Cluster analysis from Nicoli, Pavon, Grayson, Emerson, Cortelazzo, and Mitra (2023)

And what do the multi-level quantitative linguistic analysis, unsupervised machine-learning methods, and cluster analysis show? “[T]hat the users’ stylistic signature is detectable alongside that of facilitators.”

The authors’ conclusion? The texts are co-authored: co-created by both the facilitatee and their facilitator:

The results clearly show that texts written via FC generally present two linguistic imprints: the user and the facilitator. Based on this result, FC is better described as a co-creation process in which two distinct and active participants collaborate in the production of linguistic content.

Here the authors acknowledge, but claim to rule out, the obvious alternative explanation: namely, that the facilitator is subconsciously channeling the imagined writing style of an imagined persona that they’ve projected onto their facilitatee. In fact, this is arguably the most plausible way in which facilitators concoct the messages that they unwittingly facilitate.

But Nicoli, Pavon, Grayson, Emerson, Cortelazzo, and Mitra suggest that this simply isn’t possible, especially given that one of the facilitators in the study facilitated 10 different individuals (the other two involved in the study facilitated six and four respectively). They seem to think that facilitators are somehow facilitating these multiple individuals “simultaneously” and that there’s a limit on how many different personas one can imagine and subconsciously channel. They also seem to think that facilitators are incapable of coming up with different styles on their own and instead would need to have “external models”:

[N]o studies have so far demonstrated the possibility of imitating more than one style simultaneously. This hypothesis, while logically possible, requires demonstration. Besides, even if we assume that facilitators consciously manipulate their style such that statistical models end up attributing texts to users instead, and also that facilitators can maintain up to ten different stylistic systems and use them in the correct contexts, we must contend with the issue of stylistic models…. However, in the case of FC, facilitators have no external models to which they can refer, so it is unclear on which basis they would organize their imitation.

Novelists, among other people, would beg to differ. Would Nicoli, Pavon, Grayson, Emerson, Cortelazzo, and Mitra admit, say, The Moonstone or Dracula or The Color Purple as demonstrations of the possibility of imitating more than one person’s (character’s) style?

Confident that they’ve dispensed with the alternative explanation for their results—facilitators channeling the imagined personas of their facilitatees—the authors draw further conclusions:

• “FC users possess a level of literacy skills, which any AAC intervention should seek to nurture and develop to improve users' quality of life.”

• “As the user is clearly a participant in text generation, there is scope for touch-based assistance to serve as a scaffold in DD individuals' linguistic development, and to contribute positively to their quality of life and connection with carers.”

• “Whether the individual does, or could develop to, generate typed text independently should not determine the value of practising and better understanding touch-assisted typing techniques.”

The authors seem unconcerned with the disturbing implications of life-long dependence on others for communication—especially communication that must be co-authored by those others. Such dependence, fortunately, is bypassed by most legitimate forms of AAC.

But the authors, like other FC proponents, repeatedly characterize FC as a form of AAC: one specifically for those with a combination of sensorimotor and cognitive disabilities. Referencing their other article (Nicoli, Pavon, Grayson, Emerson, and Mitra, 2023), the authors suggest that the issue for such doubly-disabled individuals isn’t a lack of linguistic skills, but of challenges involving cognitive overload and planning difficulties that the facilitator’s contact on the facilitatee’s arm or torso somehow ameliorates. Linguistic skills, purportedly, are intact, and facilitation serves to unlock them, providing

cognitive simplification by reducing the cognitive demands of formulating a full-language expression… or by reducing the overall cognitive demands of the task by supporting attention, providing emotional reassurance and reducing the cognitive demands of the co-occurrent motor task while maintaining text-based communication…

As I’ve discussed in earlier posts, for example here, this is at odds with decades of rigorous research on the autistic minimal-to-non-speakers who constitute the majority of people subjected to FC.

Second, the authors cast FC as a form of scaffolding. But scaffolding, by definition, is something that (1) supports something—i.e., existing skills and (2) will eventually be removed—i.e., when existing skills reach mastery.

Third, they characterize FC as a teaching aid. But they don’t say what it’s teaching. What could it possibly be that facilitators are instructing the facilitatees by touching their arms or torsos?

Then there’s the problematic notion of co-authored messages: messages, that is, that are purportedly co-constructed by the facilitator and facilitatee. On one hand, the authors acknowledge that who constructs what is unclear from their data/results:

The nature of the corpora and these methods do not allow us to discern the details of what originates with the user and the facilitator. The scope of these analyses is at the stylistic level with no presumption nor power to distinguish facilitators' traits from users' features on a sentence-by-sentence basis. Also, these analyses do not allow any comment on how the facilitator influences and supports the user. These questions require other methods and task-oriented analyses. What the present results clearly show is that the user is not linguistically passive, and the facilitator moulds rather than wholly constructs the typed text.

In addition, they acknowledge that:

Co-authorship does not offer certainty that a message written by integrating two linguistic sources fully mirrors the FC user's own intention.

But then, noting that what was most stylistically distinct about the facilitated messages of given facilitatees were their unusual word choices, they hypothesize that the facilitatees supplied the content words and the facilitators supplied the syntactic and morphological framework:

[The] facilitators create syntactical structures within which users can fill in their own content… Syntactical scaffolding by the facilitator may also involve adjusting morphological endings, suggesting linkers or auxiliary verbs, or providing syntagmatic prompts to help begin communication (e.g. “I think that …”, or “I feel that …”)…. In this respect, the facilitators' work would be comparable to that of editors.

How one person supporting the arm or torso of another person promotes this kind of division of linguistic labor (content words vs. syntax), editing (adding and adjusting of endings and function words), and syntagmatic prompting is left unsaid. Such a linguistic cooperation/co-construction, if that’s actually what’s going on, is unlike anything seen in any other interpersonal collaboration anywhere else on planet Earth.

What’s much more likely is that the collaboration, if that’s the right word for it, operates at the level of individual letters. Sometimes the facilitatee manages to hit an unexpected letter, and the facilitator’s non-conscious response is to cue one of those unusual words or neologisms so often seen in FC.

Actual communicative collaborations, of course, involve conscious behaviors. Given this, if the authors were truly convinced that that was what was going on and genuinely curious about who was contributing what, why didn’t they simply query the facilitators? Why didn’t they dispense with all the quantitative cluster analyses and post-analytical speculation and simply ask them, directly: “Which parts of the sentence did you contribute and which parts did you leave up to the facilitatee?”

Better yet, talk the facilitators—and yourselves—into doing a simple, straightforward message-passing experiment.

REFERENCES

Bernardi, L., & Tuzzi, A. (2011). Analyzing written communication in AAC contexts: A statistical perspective, Augmentative and Alternative Communication, 27:3, 183-194 DOI: O.3109/07434618.2011.610353

Emerson, A., Grayson, A., & Griffiths, A. (2001). Can't or Won't? Evidence relating to authorship in Facilitated Communication. International Journal of Language and Communication Disorders, 36, Supplemental: 98-103. DOI: 10.3109/13682820109177866

Faure, P., Legou, T., and Gepner, B. (2021). Evidence of Authorship on Messages in Facilitated Communication: A Case Report Using Accelerometry. Frontiers in Psychiatry. DOI: 10.3389/fpsyt.2020.543385

Grayson, A., Emerson, A., Howard-Jones, P., and O’Neil, L. (2012). Hidden communicative competence: Case study evidence using eye-tracking and video analysis. Autism. 16 (1); 75-86. DOI: 10.1177/1362361310393260

Jaswal, V.K., Wayne, A. & Golino, H. (2020) Eye-tracking reveals agency in assisted autistic communication. Scientific Reports 10, 7882. DOI: 10.1038/s41598-020-64553-9

Nicoli G, Pavon G, Grayson A, Emerson A and Mitra S (2023) Touch may reduce cognitive load during assisted typing by individuals with developmental disabilities. Frontiers in Integrative Neuroscience 17:1181025. doi: 10.3389/fnint.2023.1181025

Nicoli G, Pavon G, Grayson A, Emerson A, Cortelazzo M and Mitra S (2023) Individuals with developmental disabilities make their own stylistic contributions to text written with physical facilitation. Frontiers in Child and Adolescent Psychiatry 2:1182884. doi: 10.3389/frcha.2023.1182884

Niemi, J., & Kärnä-Lin, E. (2002). Grammar and Lexicon in Facilitated Communication: A Linguistic Authorship Analysis of a Finnish Case, Mental Retardation, 40:5, 347-357 DOI: 10.1352/0047-6765(2002)040<0347:GALIFC>2.0.CO;2

Saloviita T. (2018). Does linguistic analysis confirm the validity of facilitated communication? Focus Autism and Other Developmental Disabilities, 33(2):91–9. doi: 10.1177/1088357616646075

Tuzzi, A., Cemin, M., & Castagna, M. (2004). "Moved deeply I am" Autistic language in texts produced with FC, Journées internationales d’Analyse statistique des Données Textuelles, 1097-1105

Tuzzi, A. (2009). Grammar and lexicon in individuals with autism: A quantitative analysis of a large Italian corpus, Intellectual and Developmental Disabilities, 47:5, 373-385 DOI: 10.1352/1934-9556-47.5.373