Paul Brocklehurst presents findings from his PhD study on stammering and covert error repair.
When we speak, sometimes the words that come out of our mouths do not correspond entirely with what we intend to say. Irrespective of whether or not one stammers, there are many things that can go wrong. We might pronounce words incorrectly; they might come out in the wrong order or with the wrong emphasis; or we might select the wrong words altogether, which can get us into trouble. An important skill that we learn in early childhood is to monitor what we are about to say and correct any mistakes before we start to speak out loud. In psycholinguistics, this process of pre-articulatory editing is referred to as ‘covert error repair’.
In 1993, two Dutch psycholinguists, Albert Postma and Herman Kolk, who had been studying the relationship between covert error repair and dysfluency, published a new theory of stammering called the ‘Covert Repair Hypothesis’. This proposed that people who stammer have error-prone language-formulation systems that cause them to make many more speech errors than normally-fluent speakers, and it is their attempts to covertly repair all these errors that cause the primary symptoms of stammering (repetitions, prolongations and blocks). In other words, stammering is a side-effect of excessive covert error repair activity.
“Failing to distinguish between errors and dysfluencies may then partially account for why stammering persists in some people despite their speech being relatively error-free.”
Although I never really considered that my speech contained an inordinately large number of ‘errors’, I never felt entirely confident that what would come out of my mouth would correspond very closely to what I intended. And, like many who stammer, my instinctive response, if I found myself doubting how well my words would come out, was always to make an extra effort to speak more clearly and accurately. The Covert Repair Hypothesis suggested to me that I may have been doing exactly the wrong thing, and that to reduce the severity of my stammering, rather than trying to speak better, I probably needed to lower my standards and stop trying to repair my speech errors.
Although the basic idea behind the hypothesis is appealing, subsequent research has failed to uncover any convincing evidence to support its underlying presumption that the inner speech of people who stammer is particularly error-prone. This is perhaps unsurprising, because, unlike the speech errors we make out loud, our covert (inner speech) errors are very difficult to measure. Currently, the only way we have of doing so is to rely on what experimental participants tell us, and, of course, we can never be sure how accurate their self-reports really are.
As part my PhD, I, with my team, decided to take on this challenge and try and design an experiment that could reliably compare the frequency with which people who stammer and matched controls make speech errors – both in their inner speech and when speaking aloud. I recruited 32 people who stammer (with help from BSA and regional self-help groups) and a corresponding group of 32 normally-fluent speakers, matched for age, gender and education. To ensure participants made enough speech errors for us to get meaningful results, we asked each of them to recite a variety of tongue-twisters, over and over again. To make sure that they all spoke at the same speed and didn’t stammer during the experiment, we asked them to say the tongue-twisters in time to a metronome, at a rate of two words per second. Half of the tongue-twisters had to be recited just using inner speech (in their minds) and the other half aloud. Each time a participant noticed themself making an error, they had to immediately type into a computer exactly what that error was.
The results of the experiment were clear. The people who stammer self-reported approximately twice as many speech errors as the controls - not only when reciting tongue-twisters aloud, but also just using inner speech. Moreover, when we later compared their self-reports to recordings we had taken of them, it was evident both groups were equally vigilant at self-reporting the errors they made.
At first sight, our findings seemed to support the Covert Repair Hypothesis. However, a few things didn’t quite add up. First of all, although the stammering group made double the errors as the control group, the overall number of errors they made while reciting tongue-twisters was still very small, and certainly not enough to account for the high frequency of stammers in their real-life speech. Secondly, there was a lot of overlap between the two groups, with some of the participants who stammered actually producing fewer speech errors while reciting than some of the controls. Furthermore, the fact that both groups were equally vigilant at self-reporting errors seemed to rule out the possibility that under-reporting would have accounted for the discrepancy between the number of errors participants self-reported and how dysfluent their everyday speech was.
After reading around the role of anticipation in stammering, we eventually found ourselves considering another explanation for the discrepancy between participants’ performances and the severity of their dysfluencies in real-life speech: that perhaps in real life, people who stammer are dysfluent not only when they make a speech error and try to repair it, but also when they anticipate that they might make an error, and then try to avoid making it. Our tongue-twister experiment hadn’t investigated this possibility because participants were only asked to self-report speech errors after they had occurred, not before.
To test the possibility that stammering occurs when people who stammer try to avoid anticipated speech errors, we devised a new experiment. For this, we used speech-recognition software to trick participants into anticipating that they might not be saying some words well enough. We asked them to say words into a microphone with the software giving instant feedback, showing them what words it had recognised. However, the software misled the participants - we had programmed it to consistently misrecognise certain words, even when they had said those words correctly. Each participant had four attempts at each word and, as predicted, when they consistently received feedback that led them to believe they were saying a word incorrectly, the frequency with which they stammered on subsequent attempts at that word increased – even when in reality they had said it correctly. So, contrary to what Postma and Kolk hypothesised, our findings seemed to suggest that at least a proportion of the dysfluencies that people who stammer produce result from their anticipating that they are not going to say a word correctly or accurately enough for it to be recognised by whoever (or whatever) they are talking to.
Although the conclusions from our speech-recognition experiment are only tentative and further experiments are needed to confirm them, if they prove to be correct, and stammered dysfluencies can result from trying to speak more clearly and accurately and from trying to avoid anticipated speech errors, it could have important implications for therapy. The key issue here is that much of what we anticipate is based on past experiences which may not accurately reflect our present abilities. This is especially true of the speech errors we anticipate, because the neuronal pathways responsible for speech and language production gradually become more developed and better regulated as we grow older. If, as children, we were highly prone to making speech errors, it would not be surprising to find ourselves doubting that our speech will be good, accurate or clear enough, and, at such times, we may make extra effort to avoid anticipated speech errors. Once established, such error-avoidance behaviour is likely to continue irrespective of our current speaking abilities, and it is therefore possible that we engage in a lot of error-avoidance that is completely unnecessary. Irrespective of whether it is necessary or not, excessive error-avoidance will cause our speech to be excessively dysfluent, and we may then start to consider our dysfluencies to be errors and try to avoid them too. If we do, a vicious circle is likely to become established, whereby dysfluencies are caused by trying to avoid dysfluencies. Failing to distinguish between errors and dysfluencies may then partially account for why stammering persists in some people despite their speech being relatively error-free. It may also explain why some of our participants who stammered actually made fewer speech errors than some of our normally-fluent controls.
From my own perspective as a person who stammers, I have learned some valuable lessons from this research. The first is the importance of distinguishing between speech errors and dysfluencies. Up until encountering the Covert Repair Hypothesis, I had indeed considered dysfluencies to be a type of speech error and did all I could to avoid them. Another lesson has been the realisation that most speech errors are trivial. They rarely have much effect on how well people understand us. Occasionally it may be necessary to avoid or repair them, but, on the whole, there is no need to do so. If we avoid speech errors every time we anticipate them, we deny ourselves the learning experiences needed in order to realise quite how trivial they are. So now, I happily allow myself to make speech errors and only repair them if it is absolutely necessary to do so. Adopting this new approach to speaking has been a profoundly liberating experience.
Perhaps my most valuable lesson has been that, when it comes to speech, the things we anticipate are often simply wrong. Despite the benefits that the ability to anticipate brings us, the reality is that past patterns don’t always repeat, and this is especially the case with speech, language and communication. Most of the speech errors that we made as children we no longer make as adults, and even if we do make them, their impact on listeners (and consequently on ourselves) is likely to be very different to how it was as children. This is good news, because it means that, contrary to much of the received wisdom about stammering, it is quite likely that as we get older, our ability to derive lasting benefit from speech therapy may increase. This has certainly been the case for me, and is something that I would never have anticipated.
This article is loosely based on ‘Revisiting Bloodstein's Anticipatory Struggle Hypothesis from a Psycholinguistic Perspective: a Variable Release Threshold Hypothesis of Stammering’ (2013) by Paul Brocklehurst, Robin Lickley and Martin Corley, Journal of Communication Disorders, 46(3), 217-237.
Paul Brocklehurst is Director of the Stammering Self-Empowerment Programme, based in Macclesfield, www.stammeringresearch.org.
 Much of this improvement is likely associated with the ongoing increase in brain white matter and decrease in dopamine production that continues right up to puberty. Both of these changes may reduce the propensity to make speech errors. In addition, the decrease in dopamine may help our sensitivity to such errors.
From Speaking Out Winter 2013 p18-19