A Theory of Stuttering
A Theory of Stuttering
The term “altered auditory feedback” is commonly used for “the electronic manipulation of an individual’s speech signal such that the individual perceives his or her speech differently in some way.” (Lincoln et al., 2010, p. 1122). However, this definition is somewhat narrow. If “auditory feedback” means that we hear our own voice and our words when speaking, then auditory feedback can be altered in many ways, for instance, by room acoustics (we hear ourselves delayed, with an echo, when speaking in a large cathedral) or by speaking in a higher or lower pitch or in an unfamiliar dialect. Such naturally altered auditory feedback often reduces stuttering (see Bloodstein & Bernstein Ratner, 2008). Furthermore, choral speech, metronome-timed speech, shadowing, noise, and other FCs in which a second auditory signal is presented alter what a person hears while speaking.
There is no fundamental difference between technically altered auditory feedback and “natural” FCs. Howell (2004) noted about delayed auditory feedback (DAF) and frequency-altered auditory feedback (FAF): “The former creates a speaking situation like that in an echoey auditorium, and the latter gives the speaker the impression of speaking at the same time as another speaker (either one with a deeper voice or one with a higher voice, depending on which way the speech spectrum is shifted)” (p. 31). DAF can further be interpreted as “reversed” shadowing (which reduced stuttering to a similar degree as “normal” shadowing in an experiment by Hudock, 2012; see Section 3.5).
A simple way to alter auditory feedback is amplification. Harris (1955) reported an experiment in which stutterers read aloud (A) in their habitual manner, without earphones; (B) with auditory feedback through earphones at a “comfortable” loudness level; and (C) with auditory feedback through earphones 20 dB above their comfortable loudness level. The number of stuttered words was reduced on average in condition B as compared with A and also in condition C as compared with B (cf. Martin et al., 1984).
Likewise, Ham and Steer (1967) reported reduced stuttering with amplified auditory feedback through earphones, but only with amplification of 60 dB and 75 dB above the participants’ natural loudness. Martin et al. (1984) found a reduction of stuttering by amplified auditory feedback only after, in a preceding experimental condition, auditory feedback had been masked by noise. More recently, Fiorin et al. (2021) found that amplified auditory feedback (65–90 dB) reduced stuttering in children and adolescents more than DAF and white noise did (see this blog entry). .
Auditory feedback through earphones alone is a kind of altered auditory feedback, since the impact of environmental acoustics on auditory perception is excluded. MacLaren (1960, p. 458) noted: “Headphones help to cut out distractions, give speech a more direct impact on the attention.” Unger, Glück, and Cholewa (2012) and Foundas et al. (2013) examined the efficacy of wearable speech aids that generate DAF and FAF. In both studies, a significant reduction in stuttering frequency was also found in a control condition, in which stutterers heard their voice with the device, that is, through earphones, but with DAF and FAF switched off.
Stuart et al. (1997) found a significant reduction in stuttering frequency when adults who stutter were speaking into a passive resonator—a toy shaped like a mic that generated an echo mechanically (the “Echo Mic”, see photo). Pelczarski and Hoag (2020) found a significant reduction in stuttering when adults who stutter were speaking in a resonant voice, that is, in “an easy, but strong, clear voice that can be heard over a distance as well as in background noise.” In both cases, participants heard their voice in an altered manner.
The four examples—speaking in a resonant voice, speaking into a passive resonator, amplified auditory feedback, and the “headphone effect”—illustrate that the fluency-enhancing effect of altered auditory feedback is not restricted to DAF and FAF. This should be considered when explaining the effect of DAF and FAF on stuttering. It suggests that not a specific delay or frequency shift compensates for a specific unknown deficit in the brain; there seems to be a more general mechanism underlying the fluency-enhancing effect of altered auditory feedback.
Lee (1950) found that a delay in the auditory feedback of speech of about one syllable length (1/4–1/5 second) evoked speech disfluencies in healthy individuals. Repetitions (not rarely at the ends of multisyllabic words) and prolongations (often of vowels) occurred, but also interjections, omissions, and other kinds of speech errors (see the analyses by Fairbanks & Guttman, 1958, and by Venkatagiri, 1980). These behaviors are referred to as the Lee effect. Lee (1951) called them artificial stutter, but they markedly differ from true stuttering (Stuart et al., 2002; Wingate, 1970). Venkatagiri (1980), although emphasizing similarities between the Lee effect and stuttering, found no tensed pauses in the speech of normal fluent individuals under DAF, that is, no involuntary blocks, which are typical of stuttering.
Fairbanks and Guttman (1958) supposed that normal speakers may produce repetitions and prolongations under DAF in a spontaneous attempt to restore the time match between speech production and auditory feedback. Whatever that may be, the differences between the Lee effect and stuttering suggest different underlying mechanisms, and it is probably not correct to say DAF causes stuttering in normal speakers. However, the Lee effect indicates the relevance of auditory feedback for speech control in normally fluent individuals.
Nessel (1958) found a fluency-enhancing effect of DAF in stutterers, which was confirmed in many later studies (see Bloodstein & Bernstein Ratner, 2008). Initially, DAF was considered primarily as a tool for making stutterers slow down their speech rate. To this end, long delay times of 200 ms or more were used, and the improved fluency was interpreted as the result of slow, prolonged speech (e.g., Curlee & Perkins, 1969; Ham & Steer, 1967; Langová et al., 1970).
However, Lotzmann (1961) and Webster, Schuhmacher, & Lubker (1970) showed that shorter delay times (50–100 ms) supported the fluency in stutterers more than longer ones. Kalinowski et al. (1993, 1996) and MacLeod et al. (1995) instructed stutterers to speak at a fast rate under DAF and found that the DAF effect on stuttering is independent of rate and thus not an effect of slowed speech (see also the theoretical reflections by Stuart & Kalinowski, 1996).
Stuart et al. (2002) hypothesized that, similar to choral speech, DAF increases the activity of the left auditory cortex in stutterers. In fact, Watkins et al. (2008) found increased activation in auditory association areas in both, stutterers and normally fluent controls, with DAF (200 ms delay) compared to speaking with non-delayed auditory feedback.
Two neuroimaging studies of the DAF effect were conducted with normally fluent speakers. Hashimoto and Sakai (2003) found greater activation in auditory association areas during speaking under DAF (200 ms delay). Takaso et al. (2010) tested three different delays, 50 ms, 125 ms, and 200 ms, and found increasing auditory activation with delay time. In an EEG study, Kittilstved et al. (2018) investigated the effect of DAF and choral speech on brain activity with normal speakers. They interpret their results as reflecting increased feedback control in both FCs.
Daliri and Max (2015a) found that the amplitude of the N1 component of the auditory-evoked potential elicited by a probe tone just before speech onset was consistently attenuated in fluent speakers. This “pre-speech auditory modulation” (PSAM) was greatly reduced or absent in adults who stutter. Max and Daliri (2019) hypothesized that PSAM reflects “neural processes involved in priming and selectively biasing the auditory system for its role in monitoring auditory feedback during speech production” (p. 3075). Daliri and Max (2018) found that DAF increased, that is, normalized PSAM in stutterers. Prior to speech onset, when PSAM was measured, participants already knew if they would speak with normal auditory feedback or DAF. The expectation of DAF probably drew their attention more to the auditory channel than the expectation of normal auditory feedback.
Together, these findings suggest that DAF works similarly to what choral reading, metronome-paced speech, or singing do, namely, by drawing the speaker’s attention to auditory feedback because it sounds unfamiliar and odd. This improves the processing of auditory feedback and its use in speech-motor control (which is hardly impaired by a delay of 50–100 ms).
Picoloto et al. (2017) investigated the DAF effect on the speech of stutterers with and without auditory processing disorders (APD). They found that stutterers without APD, as a group, significantly benefited from DAF (100 ms delay), but stutterers with APD, as a group, did not, possibly because their tendency to ignore apparently irrelevant auditory input was stronger than the attraction by the DAF.
The neuroimaging results mentioned above show a similar DAF effect on brain activation in stutterers and fluent speakers. Hence, the question arises: Why does DAF reduce stuttering in stutterers but cause disfluencies in normal speakers? To account for this apparent paradox, we must consider the different delay times used for stuttering reduction and for evoking the Lee effect.
The delay time best for reducing stuttering is 50–75 ms (Armson & Kiefte, 2008; Kalinowski et al., 1996; Lincoln, Packman, & Onslow, 2006). Auditory feedback with such a short delay seems to be still useful for the monitoring system without problem; it hardly evokes disfluencies in normal speakers (Foundas et al., 2004a, 2013; Stuart et al., 2002). Even a delay of 120 ms has little effect on normal speakers (Foundas et al., 2004a).
The delay time maximally disrupting for normal speakers is about 200 ms (Fairbanks & Guttman, 1958; Stuart et al., 2002). Such a long delay probably disturbs self-monitoring and affects audio-phonatory coupling. This causes the Lee effect, not only in normal speakers, but also in stuterers, which manifests in prolonged speech and normal, not stuttering-like errors such as interjections (Neelley, 1961; Hayden, Scott, & Addicott, 1977 ).
FAF means that speakers hear their own voice through headphones higher or lower than normal, usually 1/4 to 1 octave up or down (Lincoln, Packman, & Onslow, 2006). Howell, El-Yaniv, and Powell (1987) discovered that FAF reduced stuttering. Previously, Ham and Steer (1967) had found that stutterers were more fluent with frequency-filtered auditory feedback, which also makes the voice sound higher or lower but is associated with a loss of distinctness.
A perceptible frequency shift is necessary to reduce stuttering, but beyond that, the effect of FAF does not depend on a specific degree or direction of frequency shift; there is no significant difference in this respect between small (1/4 octave) and larger shifts or between shifts up- and downward (Hargrave et al., 1994; Stuart et al., 1996).
FAF can be interpreted as a simulation of choral speech (Howell, 2004; Kiefte & Armson, 2008). Accordingly, Ingham et al. (1997), referring to the PET study by Fox et al. (1996), supposed that if choral reading increases the activation in auditory association areas in stutterers up to a normal level, then FAF may do the same. In fact, Watkins et al. (2008) found greater activation in the bilateral superior temporal cortex in both, stutterers and normally fluent controls, during reading with FAF (half an octave upwards) compared to reading with unaltered auditory feedback.
This is consistent with the view that FAF increases attention to auditory feedback because it sounds odd. However, FAF and choral speech differ in efficacy. Kiefte and Armson (2008) found that all of their 12 participants exhibited a stable pattern of less than 2% stuttering with choral reading, but only 3 of the 12 participants exhibited a similarly strong and stable effect when reading under FAF. The authors report an average reduction of stuttering by choral reading and FAF of 90–100% and 60–90%, respectively. The difference may result from the fact that listening to auditory feedback is task-relevant in choral reading, but not in speaking under FAF.
The fluency-enhancing effect of FAF decreases with time through familiarization. Armson and Stuart (1998) reported that 6 of their 9 participants who benefited from FAF showed approximately 50% reduction in stuttering at the beginning of a reading task with FAF, but within 10 minutes, percent stuttering returned to the baseline values assessed with natural auditory feedback.
A further factor modulating the fluency-enhancing effect of FAF is the amount of attention needed for speech formulation. The reduction of stuttering by FAF is smaller on average during self-formulated speech than during reading; some stutterers do not benefit at all from FAF when speaking spontaneously (Armson & Stuart, 1998; Ingham et al., 1997). Together, the overall smaller and more inconsistent effect of FAF on stuttering is not surprising, given that it is an effect on attention by an irrelevant stimulus.
Howell, Sackin, and Williams (1999) found the reduction of stuttering by FAF during reading to be greater in adults than in children who stutter. The authors proposed an explanation, whose core is that the FAF effect on stuttering results from a reduction in rate. But Kalinowski et al. (1993) and MacLeod et al. (1995) had already demonstrated that the FAF effect does not depend on a reduced rate. A different efficacy of FAF in children and adults is less surprising when the FAF effect is considered as an effect on attention. In addition to the FAF, attention is influenced by other factors, which may be different in children and adults, e.g., the amount of attention needed for reading, the lab environment, or the headphone effect (see above).
FAF and DAF have often been combined in electronic speech aids. However, MacLeod et al. (1995), who tested the efficacy of this combination in the laboratory, did not find a greater reduction of stuttering than with DAF or FAF alone. This is not surprising if both work similarly, drawing the speaker’s attention to auditory feedback as much as possible, depending on the person and situation. DAF and FAF cannot do more together than alone in this respect. But their combination makes sense given (1) individual differences in the sensitivity to DAF and FAF and (2) the problem of familiarization, which can be reduced by changing DAF/FAF settings.
At the end of the section, I include an old, short video in memory of Andrew Stuart and Joe Kalinowski, who gave us a lot of knowledge about the effect of altered auditory feedback on stuttering, and who developed the SpeechEasy—a device that should not be used as a kind of prosthesis, but rather as an aid to listen to auditory feedback when speaking.
Pay attention to what happens when Joe puts the device in his ear and starts talking. He begins fluently—stuttering rarely occurs immediately at speech onset, where auditory feedback is not yet available. Then a block occurs—auditory feedback is now provided but probably poorly processed because of insufficient attention to it. Finally, the device begins to work, draws attention to the altered auditory feedback (which is, despite DAF and FAF, still usable for speech control), and the stuttering disappears.