Mediated Communication

In this paper, I will examine Facilitated Communication (FC), a method in which people who lack functional speech, usually due to a developmental disability, type with the assistance of a facilitator. My goal will be to determine whether this method is effective, and whether it is ethical. Two things distinguish my inquiry. First, due to the population of individuals involved, FC has the potential to be both vitally important and deeply problematic. People with whom FC is tried tend to have few communicative options. They may speak or sign a few words, but they are essentially nonverbal and most have been diagnosed as intellectually disabled.1 Thus, if FC works, it could provide a valuable connection, enabling them to interact with the rest of the social world. If FC doesn't work, its negative impact falls on a uniquely vulnerable group of people. Second, my approach to the problem is that of analytic philosophy. Fairly comprehensive literature reviews, by scholars better versed in empirical methods than I, have already been published. While including key points from the empirical research, my emphasis will be on analyzing and evaluating the arguments surrounding FC.

Facilitated Communication is, at least ostensibly, a form of mediated communication. So, in this first section, I will begin by considering mediated communication generally, answering the following questions: What is mediated communication? What factors may make mediated communication problematic? With this conceptual framework in place, I will better be able to address the specific challenges that arise with regard to FC.

Pavlik and McIntosh define mediated communication simply as "the process by which a message, or communication, is transmitted via some form, or medium" (70). But doesn't all communication take place through some medium or other? Lacking telepathy, we use words, signs, pictures, etc., as mediums to convey ideas and emotions from person to person. One might express oneself by making a series of hand motions or by emitting a succession of sounds — that is, through different mediums, which are modulated in such as way as to convey language. When communication is called mediated, however, that connotes at least a second layer of mediation, "employing an external medium between sender and receiver" (Fang 7).

Thus, if I use only my own body to produce a message, and a companion observes me and understands the message, this would be direct, unmediated communication. While I employ a linguistic medium (say, speech), there is nothing additional mediating between my companion and me. However, if I use a tool, such as a pencil, cell phone, or computer, or if there is some other agent involved, such as an interpreter, this is mediated communication. Mediated communication systems used by people with disabilities include communication boards (displaying pictures, symbols, or words), computers and other electronic devices that accept typed input, and pressure switches that trigger recorded words or phrases. Such systems are classified under the rubric of "augmentative and alternative communication" (AAC), though this term doesn't precisely match the concept of mediated communication. For instance, a person who uses gestures to compensate for articulation difficulties is said to be using AAC, though her communication is not mediated (Downey and Hurtig). Of course, if only her family members understand her gestures, and they interpret them for others, then the communication becomes mediated. In the case of FC, there are technically two mediators involved: the AAC device used, such as a keyboard, and the facilitator who helps the person interacts with the device.

Intuitively, it seems that some instances of communication are more mediated, or more problematically mediated, than others, but it is not obvious how best to understand this range. There is no cohesive literature to consult for guidance. As evidenced by online and library searching, most references to the general term 'mediated communication' actually target one of two sub-areas: computer mediated communication and mass media. While FC often involves computer interfacing, it is not an essential component, and FC is typically used for interpersonal communication rather than mass communication. Somewhat more relevant is literature on specific kinds of interpersonal mediated communication: language interpretation, particularly American Sign Language, and augmentative and alternative communication.

Aided by these sources, I offer the following factors for consideration as determinants of the significance of mediation upon communication. If one or more of these factors is compromised, the mediated communication is problematic. Problematic communication does not necessarily imply that there is something ethically wrong with the situation, but it serves as a clue to look for ethical issues and a framework for understanding them when they arise.

Multiple Intermediaries → Dependability

One might think that, since mediated communication requires at least one external medium, the more intermediary steps, the more significantly mediated the communication. On further reflection, however, counting the number of intermediaries is not all that helpful — we don't typically care how many satellites a radio signal has bounced off, for example. Actually, the number of steps in the communication process might matter, not because this quantity is intrinsically important, but because multiple intermediaries can diminish dependability. The more parts a communication system has, the higher the chance that something will break down. A recent study of speech generating communication devices found that more than 40% required repairs in their first year of use, with 66% of those machines breaking down a second time within the five-year study period (Shepherd et al.). The many steps (components) between user input and speech production presented multiple opportunities for failure. The dependability of the device is a determinant of whether users and those with whom they converse "buy into" an AAC system and stick with it, allowing the user to gain proficiency.

Accuracy and Neutrality

Multiple intermediate steps may also raise concerns about accuracy, though accuracy issues can impact mediated communication even when there is only one intermediary. In a review of the various iterations of the Registry of Interpreters for the Deaf's Code of Ethics, from the initial 1965 version to the 2005 Code of Professional Conduct, Stewart and Witter-Merithew identify seven core responsibilities of interpreters, two of which target accuracy: accuracy of interpretation and discretion in accepting assignments (114). The latter is relevant because a criterion for accepting assignments is having the skill to interpret successfully in the given situation. With regard to the former, the Code of Professional Conduct states that the interpreter must "render the message faithfully by conveying the content and spirit of what is being communicated, using language most readily understood by consumers, and correcting errors discreetly and expeditiously" (RID 3).

Conveying the content and spirit of a message is easier said than done. Some models of interpreting (whether ASL/spoken language or spoken language/spoken language) envision the interpreter as a conduit through which the communicator's language is converted to the recipient's language with machine-like efficiency. Wilcox and Schaffer argue that this account is simplistic and neglects the cognitive processing that takes place when one comprehends language and produces translations (39). Rather than decoding a message in one language and encoding that precise meaning in another, the interpreter must actively construct meaning, inferring it on the basis of the linguistic input. 100% accuracy may not be feasible.

Accuracy is intertwined with neutrality, as the influence of the interpreter's attitudes or opinions can skew the message conveyed. The conduit model of the previous paragraph is a reaction against an older concept of interpreter as friend and helper, which became seen as paternalistic (Conrad 94; Janzen and Korpinski 167). A friend may wish to shield one from harsh news, or soften the expression thereof, and a helper may elaborate on one's statements if she perceives this to be in one's best interest. Conrad discusses the case of a patient at a community health clinic, a recent widow and new immigrant, whose doctor tells her that she has HIV through a gruff, "Yup, you're positive" (90). The interpreter translated the doctor's statement as "the tests are positive," earning reprimands from fellow interpreters and trainers for softening the delivery. Perhaps the interpreter was trying to cushion the blow; however, since Conrad does not say which language the client spoke, it is also possible that a direct translation of "you're positive" would not make sense or convey the intended meaning in the target language. Thus, even with skillful and well-meaning intermediaries, mediation that involves another human agent can complicate communication.

Control of Output

In a classic Star Trek episode called "The Menagerie," the paralyzed Captain Pike uses a communication system with exactly two outputs: one flash of a light for yes, two for no. One would think we could do better than that by the 23rd century! In contrast, a contemporary computer based voice synthesizer system may give indefinitely many potential outputs — any possible utterance in the relevant language. Yet even this provides less flexibility than unmediated spoken language, as one lacks control over vocal inflection. More output options are not always better, however, since the physical and cognitive demands of interacting with a highly nuanced system can overwhelm the user, reducing functionality. Scholarship in Augmentative and Alternative Communication has highlighted the need to maximize users' functional control by carefully matching the AAC system to the user and involving the user and her family and peers as much as possible in modifying the system, such as through choosing or editing vocabulary options (Zangari et al.; Schlosser and Lloyd).

Another sort of control over output ties into a concern many of us have for our everyday use of mediated communication: confidentiality. Some forms of mediated communication leave enduring traces, the disposition of which the communicator does not fully control. The materials we post on social networks can be used in ways we don't sanction; our employers may read our emails. With regard to the sorts of mediated communication particular to disabled individuals, human-mediated interactions make at least one other human privy to one's information.

Access to Feedback

Suppose Andrew has an electronic communication board displaying eight symbols. He presses one, and the device announces, "Time for the bus." He hears this, grins, and then presses the next button over ("I want a drink"), then presses it again just to be clear. Access to the output of his AAC device allows him to correct his input. He may also get another kind of feedback, such as his teacher telling him, "No, it's not time yet." Ordinary communication contains multiple overlapping feedback loops among conversation partners, as does successful mediated communication (Tonsing et al. 31-2). However, in mediated communication, one does not always have access to sufficient feedback, or the opportunity to correct misunderstandings. For example, a Deaf person communicating through an interpreter has access to her audience's facial expressions, but not the specific spoken words they heard.

To sum up this section, dependability, accuracy, neutrality, control of output, and access to feedback determine how significant mediation is for an instance of communication. The relationship between these factors and ethical considerations is not straightforward. Not every mediated communication glitch is evidence of wrongdoing, but they can serve as red flags for further consideration.

History of Facilitated Communication

With the above factors in mind, let's examine a particular form of mediated communication that has generated considerable controversy: Facilitated Communication. In FC, a facilitator provides physical and emotional support for a person with communication difficulties, enabling him to type on a keyboard or point to letters on a communication board. The degree of support varies, from holding the person's typing/pointing hand and isolating his index finger, to holding the wrist, forearm, or sleeve, to cupping the person's elbow or keeping a hand on his shoulder or back. Proponents of FC encourage the fading of physical support over time, if possible, though they regard the facilitated subject as actively communicating even under more intense support conditions.

Rosemary Crossley developed FC in Australia, in the 1970s, using it first with people diagnosed with cerebral palsy. Right from the start, there were questions about who truly authored the messages, the putative communicator or the facilitator. However, when one of Crossley's first clients, Annie McDonald, used FC to sue for release from the institution in which she lived, the Supreme Court of Victoria concluded that McDonald was indeed the source of the statements she produced with facilitation. Among other things, McDonald was able to spell out, via FC, words that she was given when her facilitator was not present (Biklen 1993, 5). Although, while institutionalized, McDonald was classified as intellectually disabled ("mentally retarded"), her performance on a court-ordered Peabody Picture Vocabulary test, taken without facilitation, showed her intelligence to be at least average (Biklen and Cardinal 13). Thus, there was no reason to believe that her facilitated output was inconsistent with her documented intellectual level, and there was an obvious physical reason why she might need help with pointing at letters, i.e., her cerebral palsy.

In the 1980s, Crossley expanded her use of FC to people with autism and other developmental disabilities. In 1988 — 1989, Douglas Biklen learned FC methodology from Crossley and brought it back with him to the United States, implementing it in school districts surrounding Syracuse, New York (Biklen 1993). Although the method tends to be associated with autism, Biklen recommended it more broadly for "individuals with autism or related developmental disabilities who do not speak or whose speech is not very functional" and "individuals who have these problems with speech and who may be presumed to have severe intellectual disabilities." He also notes that FC may be appropriate for individuals with speech problems and physical difficulties such as problems with eye-hand coordination, pointing with an index finger, and muscle tone (1993, 195).

FC became popular in the first half of the 1990s, sparking hope and enthusiasm among parents and teachers of individuals with autism (Dayan and Minnes). One can see how such hope was needed, given the dismal outlook for autistic individuals proposed at the time. For example, in 1990, Seifert wrote:

Despite therapeutic intervention, the autistic child fails to develop reciprocal social interactions, communication skills, and imaginative activity … Occasional reports of an autistic adult marrying, having children, and living a successful life will always be found to reflect either an incorrect diagnosis or a variation of the disorder. (5, 7)

Biklen, on the other hand, exhorts us to assume competence and communicative potential, and reports that, for many of the students who tried FC, "words and sentences seemed to virtually explode out of them; we were there to catch and admire them" (1993, 60).

Concerns over the validity of FC attained urgency when several individuals, via facilitation, made allegations of sexual abuse. Some of these reports were corroborated by physical and other evidence; others, after investigation, remained unsubstantiated (Botash et al.). These cases prompted testing to determine who was truly the source of the accusations: the developmentally disabled individual or the facilitator. The tests typically failed to validate FC. For example, in an evaluation of one such case, Hudson et al. found that when the subject and facilitator heard the same questions, facilitated responses were correct 80% of the time. When subject and facilitator heard different questions, through headphones, none of the facilitated responses correctly answered the subject's questions, while 40% of the facilitated responses correctly answered the facilitator's questions. Investigation spread beyond court cases, with many more studies challenging the efficacy of FC.

The popularity of FC has waned considerably since the 1990s, though it is still embraced by pockets of enthusiastic supporters. The subject is polarizing. It has been dubbed "pseudoscience" by some researchers (Finn et al.), and prominent disability advocate Wolf Wolfensberger compared it to the cold fusion hoax in nuclear physics. Professional organizations, such as the American Speech-Language-Hearing Association, the Association for Behavior Analysis International, and the American Academy of Child and Adolescent Psychiatry, regard FC as discredited. Yet Biklen, now Dean of the School of Education at Syracuse University, continues to endorse it, and the Syracuse University Institute on Communication and Inclusion (formerly the Facilitated Communication Institute) offers training and workshops on FC. There continue to be anecdotal reports from people who say that FC has radically improved their lives or the lives of their children. And some individuals who began typing with facilitation now type independently. Response by autistic self-advocates in the emerging neurodiversity movement appears mixed.2

Is Facilitated Communication Valid?

In everyday English, the word 'valid' often just means something like 'correct' or 'compelling', e.g., "she made a valid point." In analytic philosophy, we use the word more restrictively, to mean an argument in which the premises, if true, logically guarantee the conclusion. In this sense, an argument can be valid; a point or statement can't be. When researchers and educators ask whether FC is valid, they are using the term in a third way, much more closely related to the term's meaning in psychological testing. A test is valid if it measures what it purports to measure. So, if a test claims to measure intelligence, but how well one does on it strongly reflects one's exposure to white, middle class life, the test lacks validity. FC is not a test, but analogously, one can ask whether the messages produced via FC represent what they are purported to represent: the thoughts or communicative intentions of the disabled individual.

Now, philosophers of a skeptical bent might pause here and note that we can't really know what thoughts or intentions are present in another person's mind, even if one can communicate directly with the other person. You can observe the other person's behavior, including verbal behavior, "but none of this will give you direct access to their experiences, thoughts, and feelings" (Nagel 20). Poststructuralists might claim that the notion of authorship is problematic, challenging "the humanist assumption that the subject is capable of fully representing himself/herself via the transparent medium of language" (Erevelles 19). But we philosophers bust our students for plagiarism just like any other college professor, so there must be a practical sense in which "Sam wrote this" and "No, Sam may have pressed the keys, but his roommate supplied the content" are debatable alternatives. Satisfying whatever practical standard there is, not absolute certainty, is what we are hoping for from FC. In the remainder of this section, I will analyze two debates within the literature relating to the validity of FC.

The Debate over the Abilities of Autistic People

An argument against the validity of FC can be stated as follows:

1. The messages produced by autistic individuals using FC are extremely sophisticated.
2. This contradicts widely accepted theories of autism.
3. Therefore, autistic individuals are not the true authors of these messages.

Variants of this argument can be found in Cummins and Prior (11-12), and in Green (71-72). The charge is that autistic individuals using FC didn't just converse or communicate basic information, they wrote poignant missives and engaged in debates over educational policy, showing keen awareness of others' perspectives. While they made some spelling and grammatical errors, the typical errors found in the speech of autistic people who can talk, such as I/you pronoun confusion, were notably absent. Standard theories of autism construe it as impairing language and social development. Such theories strongly suggest that this level of communication is highly unlikely.3

Biklen, naturally, objects to the jump from premise 2 to the conclusion in 3. He agrees that, via FC, autistic individuals have produced striking examples of communication, and that this does not jibe with the dominant understanding of autism — so much the worse for the dominant understanding of autism! In its place, he proposes that autistic individuals have neuromotor difficulties relating to voluntary action (apraxia); there is nothing wrong with their cognition or language competence, just difficulty with expression that FC helps them overcome (1993). I'm inclined to agree with Biklen that conflicting with dominant theories is not, alone, a good reason to discount a phenomenon. With the benefit of nearly twenty years of hindsight, we can see that these theories continue to morph and develop. While the notion of apraxia does not seem to me to be a sufficient explanation for autism spectrum disorders, that's outside my area of professional competence. I would, however, like to pose one challenge: if FC works by mitigating the apraxia of autistic people, why is it also used with people who have severe intellectual disabilities but who don't have autism?

Counter to the anti-FC argument, above, Biklen seems to reason:

1. As educators, we ought to presume communicative potential. (1993, 47)
2. Individuals diagnosed with autism and/or severe intellectual disabilities have the potential to communicate through typing.
3. The developmentally disabled individual is the genuine source of FC output.

Now, I'm not suggesting that Biklen thinks this is a valid argument; clearly it is not. Rather, he asserts all three clauses separately, sometimes bouncing back and forth between them. I present the three claims to emphasize that 1 does not imply 2, and that one may accept 1 and 2, yet reject 3. To support 3, we need data that goes beyond potential to address what actually happens in FC. This brings us to the second debate.

The Debate over the Empirical Evidence

1. Research studies that incorporate control procedures overwhelmingly conclude that FC is ineffective.
2. Studies that do not use control procedures conclude that FC is effective.
3. Studies with control procedures are more dependable than those without.
4. Therefore, FC is ineffective.

Several reviews of FC literature were compiled in the mid-1990s, two by Green, in 1992 and 1994, and three by Hudson, Jacobson et al., and Simpson and Myles, in 1995 (reported in Mostert).4 Green classified studies as those that include control procedures and those that do not, a classification repeated by later authors. As Mostert explains, "controlled" educational research does not necessarily adhere to the ideal standards of empirical testing. Assigning subjects randomly to experimental and control groups, or even having a control group at all, may be impractical. So, when Green calls some FC studies controlled, that means the research includes procedures designed to control for reasonable competing explanations of the phenomena observed (Mostert 291).

In the case of FC, the obvious competitor to the hypothesis that the developmentally disabled subjects are the authors of the messages they type is the charge that the facilitators author the messages, or at least influence them. Thus, the FC studies dubbed "with control procedures" are those that incorporate one or more methods designed to distinguish these hypotheses; the "no control procedures" studies lack such methods. Two common control methods in FC studies are stimulus screening and message passing. In stimulus screening, both subject and facilitator are present together, but some technique is used to ensure that only the subject has access to test content (words, pictures, questions) that she must respond to via FC. The facilitator is either "blind" or misinformed. In message passing, the subject sees or does something before the facilitator enters the testing room, then is asked to report on it via FC.

Green found that the 12 studies with control procedures she reviewed in 1992, and the 25 she reviewed in 1994, showed that FC is not effective. For example, putting the 1994 studies together, only 12 of 226 facilitated performances were better than one would expect from chance. In contrast, the 9 studies with no control procedures supported FC, with almost all subjects displaying significant literacy skills. In a follow-up review of FC studies published between 1995 and 2001, Mostert found similar results: 19 studies with control procedures refuted FC, while 4 studies with no control procedures supported it. However, four studies with control procedures were "vaguely supportive of FC" and two claimed to substantially support FC, though Mostert critiqued the experimental methods of these latter two studies. Thus, Green and Mostert assert premises 1 and 2, in the argument above, assume premise 3, and draw the conclusion, 4, that FC is ineffective.

Some weaknesses in the data should be noted. First, many of the studies consisted of small numbers of subjects — an "N" of less than 10 was common. Second, in a few cases, the facilitators and subjects were not experienced at FC or familiar with each other; the fact that novices can't perform well quickly is not an indictment of the communication technique itself. In addition, FC proponents have argued that, counter to premise 3, above, the testing procedures in the controlled studies may have inhibited performance (e.g., Biklen, Cardinal et al.). Unfamiliar settings, partitions, headphones, even the very idea of being tested, might make subjects loose confidence, become anxious, or otherwise refuse to perform, making the controlled studies less accurate than the studies that eschewed these complications.

Montee et al. responded to this objection by designing a controlled study that specifically addressed the concerns of FC proponents. Seven subjects participated with their usual facilitators in familiar settings. The subject/facilitator dyads were selected on the recommendation of their service providers as their most competent FC users, regularly producing conversational level outputs. The tasks were chosen to avoid any potential word-finding difficulties that might cloud the results. Message passing was used as a control method, as this is regarded as less intrusive than stimulus screening (which was also used in some of the tasks). If the subject showed signs of anxiety, the trial was terminated; facilitators also had the authority to terminate trials if they felt uncomfortable. The subjects viewed pictures and performed activities and, in various trials, their facilitators (a) observed the activity or picture, (b) did not observe the activity or picture, or (c) were given false information about the activity or picture. Then the subjects were asked what they saw or did, and responded via FC. When the facilitators observed the picture/activity, facilitated responses were 75% correct for pictures and 87% correct for activities. When the facilitators did not observe, or got false information, correct responses were minimal: 0% to 1.8%. In the false information trials, 66% (for pictures) to 80% (for activities) of the time, the facilitated output matched what the facilitator was told, not what the subject saw or did. These results are not unusual among controlled studies; what is significant is the extent to which the researchers attempted to mitigate any factors that might impede the subjects' performance. If the control procedures were indeed at fault, one would expect poor or sporadic performance, not output that matched the facilitators' information.

Against premise 1, at least one controlled study, by Cardinal et al., provides fairly compelling support for FC; this is one of the two studies critiqued by Mostert. Cardinal's study had a relatively large number of subjects, 43, and used the message passing approach. Subjects, without their facilitators, saw words taken from a pool of 100 words used in their functional reading curricula. In the baseline condition, subjects were encouraged to type the words they saw without facilitation; then, in the experimental condition, they typed with facilitation. There were no trials in which the facilitator was told the word, or a different word, so there was no direct measure of facilitator influence. The results showed that the subjects' performance improved significantly over baseline, but not right away. The mean baseline score was .11 (out of 5) words correct; the mean for the first 10 trials after baseline was .29, which improved over time to 1.60 on trials 81-90. This argues for an extended study design to maximize performance.

However, even the later scores don't reflect the striking communicative performances that are often ascribed to FC; even in later trials, on average subjects got only 32% of familiar 3 — 6 letter words correct. What interests me about this study is the potential for truly mixed messages. Facilitators had access to the 100-word pool, which gave them a none-too-helpful 1% chance of guessing their subject's word on any trial. But, if the subject got the first letter right on his own, that would boost the facilitator's odds considerably — how much depends on the number of words beginning with that letter there are in the pool (numbers range from 1 word each beginning with J, K, L, and O, to 16 words beginning with B). I'm not suggesting that the facilitators intentionally tried to scam the test, but the set-up raises interesting prospects for combined authorship of facilitated responses.

So, is FC valid? My conclusion is that it has not been shown to be. That doesn't mean it never genuinely enables people to communicate better than they would without it; based on Cardinal's study and observations by FC proponents, I believe that sometimes it does. But the preponderance of evidence shows that FC output quite often represents the facilitator's thoughts rather than the communicative intent of the disabled person. I assume that this influence is typically unintentional, which makes it hard to guard against. It would be helpful to know whether there is some level of support that reliably shows no facilitator influence. One study (Smith et al.) suggests that error resistance is the culprit in conveying facilitator influence. Consider Biklen's explanation of what facilitators do:

When facilitators work with students, watching their selection of letters, pulling them back after each selection, slowing them down, and giving verbal encouragement, often calling out letters as they are typed and saying words as they are completed, students still often hit letters next to intended letters or type echoed words or phrases as well as seemingly nonsensical sequences of letters. Thus, the facilitator must be "all eyes," working hard to hold the student back from obvious errors (e.g., perseverative strikes at the keys, frequent hits of keys next to correct choices, three or four consonants in a row). In other words, the facilitator aids the student by assisting physically, but not guiding. (1993, 123)

This makes it sound like facilitators are not supposed to push their partners toward the right choices, but they are supposed to hold them back from many of the wrong ones. Unsurprisingly, while doing this, the facilitator shapes the message. Perhaps when facilitator support is faded back to the point that the facilitator can't cue or avert errors, we can be more confident that whatever communication occurs is genuine.

Is Facilitated Communication Ethical?

Facilitated communication raises ethical concerns because several of the factors significant to mediation, described in section 1, appear to be compromised. Evidence shows facilitator influence over the content of messages, which gives good reason to question the accuracy and neutrality of FC. While FC provides flexible, nuanced outputs (anything one can type), it is not clear to what extent the putative communicator controls these outputs. While the facilitated subject has access to what has been typed, she can only correct or modify it through FC, which she may not control. (Facilitators are instructed to ignore non-facilitated statements. If someone says "No" while typing "Yes," they go with the "Yes.") Finally, suppose that FC proponents are right and testing conditions greatly impede subjects' ability to communicate via FC. That seems to show that FC is not a very dependable method; it breaks down easily.

All this appears to point to the following argument:

1. The validity of FC is not well supported; in fact, there is good reason to question it.
2. Therefore, FC is unethical.

But I think the situation is not quite that simple. The fact that something is scientifically unsupported does not directly imply that it is unethical. The complicating factor for FC is that it sometimes produces successful "graduates." That is, there are documented cases where people with diagnoses of autism and/or intellectual disability could not communicate through speech, then typed with facilitation for a period of time, then eventually progressed to typing (and sometimes speaking) independently. The writings of some of these individuals are included in Biklen, 2005. Of course, once someone is typing independently, her communication is no longer "facilitated," so the authenticity of her current communication does not prove the authenticity of FC. And one can't say for sure that these independent typists would not have achieved this skill without FC. Still from the point of view of, say, a parent deciding whether to use FC with his child, is it wrong to play the long odds? Even if the chances are low, the potential payoff is high. While one may criticize the mathematical awareness of someone who plays the lottery every week, unless he is playing using the family rent money, it's not an ethical failing.5

To support the conclusion that FC is unethical, one would need to add an additional premise stating that FC causes more harm than benefit, or that it, in some way, violates someone's rights or interests. With regard to harm/benefit analysis, one significant, but perhaps improbable, benefit has already been identified — an FC user could graduate to independent typing. Also, despite compromised accuracy and neutrality, FC users may sometimes get their ideas across. Given the fact that FC is typically tried with those who have had very little success in communicating their thoughts, even occasional success would be a boon. Even if, for a particular individual, the facilitated output is complete fiction — that is, the communication originates solely with the facilitator — social benefits might still accrue to the individual. Those around him might regard him as intellectually competent and be more inclined to interact with him. Educational opportunities may open up. Quite a few students were switched from self-contained classrooms for the intellectually disabled to mainstream placements, using FC. Even if these placements did not match the students' true achievement levels, they might be more stimulating and socially satisfying. Today, this benefit is less contingent upon FC than it was in the past, as mainstream, or rather, inclusive, placements are now more likely with or without FC, thanks in part to Biklen and others who have argued strenuously for inclusion.

As for harms, well, FC is not physically dangerous, nor is there evidence that it is painful or unpleasant for the participant. This is worth mentioning, as there are autism interventions that do cause pain (aversive therapy) and potential physical harm (e.g., chelation therapy). Social harms have accrued when false allegations of abuse led to removal of the child from the home and familial stress. Counter to this, we should include those instances when the allegations turned out to be true. Even if the victim was not really the source of the allegations in these cases, FC provided an outlet for the facilitator to unconsciously voice suspicions. In addition, FC comes with opportunity costs. Time spent communicating (or not really communicating) through FC is time spent not learning a potentially more functional AAC system or other useful skills.

The primary harm associated with FC is the significant risk that the user's thoughts or communicative intentions are being misrepresented. Granted, the person whose ideas are misrepresented may not be aware of what is going on, and thus, may not feel angry or frustrated. But even in such a case, it seems that a considerable wrong is being done or right violated. As Dayan and Minnes put it, "if a client's communications are being influenced unwittingly or otherwise, the client in effect is being used as a puppet, an act which many would see as violating the individual's respect and dignity" (185). Much of Biklen's rhetoric is aimed at those who do not credit people diagnosed with autism or intellectual disabilities with having potential, with having something to say even if they can't say it. But it is precisely by adopting the stance that such people are potential thinkers and communicators that the puppet analogy becomes most chilling. They are people, already marginalized; it matters that we don't also put words in their mouths. However, it might not be such a bad thing to misattribute communicative intent on a temporary or tentative basis. We do this all the time with young children. At 8 months of age, my godson signed his first word. OK, he opened and closed his hand in a way that somewhat resembled the sign for milk. His mother and I stand by our attribution of communicative intent. We could be wrong. That's all right. The fact that we enthusiastically reinforced his "sign" might contribute to his subsequent language development.

Thus, I conclude that uncritical, long term use of FC is unethical, as it can violate the individual's right not to be manipulated or misrepresented. However, if all decision makers involved are aware of the empirical challenges and want to implement FC on a trial basis, in the hope that it will serve as a stepping-stone to more independent communication, this may be ethically OK. There is nothing morally problematic about teaching people to communicate by typing, and hand-over-hand assistance, followed by the fading of such assistance, is a well-established teaching tool. Taken that way, FC is just another method of learning to use an AAC device, a keyboard. Where it becomes concerning is when facilitated output is presumed to be authentic even under physical support conditions that risk facilitator influence, and when facilitated output is privileged over other communicative attempts.

Attempting FC as only one part of a comprehensive AAC program should help avoid these pitfalls. Now that the "fad" phase of FC has passed, the practice seems to be shifting in this direction, with more emphasis on finding ways that a not-yet-independent typist can self-corroborate messages (e.g., Biklen 2005, 7-9). For instance, if one can read aloud what one has typed (a task some autistic people find easier than independent speech), that doesn't prove authorship, but it does show meaningful access to the mediated output. If one can also independently signal yes/no, accepting or rejecting the messages produced, that gives one some control over the accuracy of the output. The "temporary and tentative" attitude I'm suggesting may not fit well with the confident and supportive tone that FC proponents prefer. And the parties involved may face a difficult decision if, after a reasonable period of time (I won't speculate how long that is) the FC output is eloquent but still requires a great deal of physical support, and the individual has not attained the skills needed to self-corroborate.

In the big picture of disability studies, FC is fascinating because it puts in tension two key values: having a voice and not being spoken for by others. Ideally, FC should provide people who have heretofore been voiceless, and thus deprived of power, with the ability to speak for themselves. Yet, as Green notes, it risks "allowing others to impose their own wishes, fears, hopes, and agendas on nonspeaking individuals" (71).