Before I further discuss the role of white matter tracts in speaking and stuttering in the next section, it may be useful to wonder how short-term memory works, and whether there is a relationship between memory and attention. There are some empirical findings suggesting that not only a deficit in attention regulation (see Section 3.3) but also a deficit in working memory plays a role in stuttering (read more). First, it must be emphasized that memory is not a location, not a particular storage in the brain. Memory is rather a function – probably a function of the brain as a whole (read more).
Second, I start from the position that the brain is an analogously, not a digitally working information-processing system. That means there is neither data transmission nor data storage in the brain. Nerve fibers transfer only excitations (action potentials) that can excite or inhibit the receiving neurons – but nothing is encoded in the action potentials. The only information contained in an action potential is where it comes from (see below). Synapses determine whether an arriving potential excites or inhibits the receiving neuron, and how strong it impacts, that is, how many action potentials must arrive at the same time for making the receiving neuron ‘fire’. The adjustment of synapses can alter by learning processes.
How can an information be kept in memory for some seconds under these preconditions? The information cannot be resided in a storage until it is needed, but must be ‘held active’ in the way that the neurons representing the information continue to fire. What does it mean: a neuron represents an information? For example, a neuron that fires if and only if the ear perceives sound waves of a particular frequency represents this frequency – let’s call it a ‘frequency neuron’. The vowel /a/ consists of several frequencies (undertone and overtones). If you hear the vowel /a/, several frequency neurons fire at the same time. A neuron that is excited just by that and only by that, represents the vowel /a/ – it fires if and only if the vowel /a/ is perceived – hence we can call it a ‘phoneme neuron’. The word /cat/ consists of the phonemes /c/, /a/, and /t/ in a certain temporal order. A neuron that is excited by the phoneme neurons representing these three phonemes and that fires if and only if the phoneme neurons fire in the order /c/–/a/–/t/ represents the word /cat/, and we can call it a ‘word neuron’ (precisely said, it represents the acoustic word form – see Section 1.2).
It should be more clear now what I meant by the above statement that the only information contained in an action potential is where it comes from: An action potential coming from the neuron representing the word /cat/ is an information that this word was just heard or remembered. Of course, a mental representation by a single neuron is unlikely. Important representations like phonemes or familiar words might be realized by many neurons, perhaps working in parallel – otherwise the death of a cell would result in a speech sound or word to forget. By the way, the semantic meaning of the word /cat/ is realized by links of the word neuron(s) of /cat/ to neurons representing visual perceptions of cats, acoustic perceptions of meowing, etc. Whenever the word /cat/ is activated, those non-verbal representations are co-activated, thus we understand the word we hear.
Let’s go back to our topic: For keeping an information in memory for some seconds, the information must be ‘held active’ in the way that the neurons representing the information continue to fire – so I wrote above. However, after sufficient excitation, a neuron can loose only one action potential – then it needs a further sufficient excitation to elicit the next action potential after a break (refractory period). What can the neuron make continuously fire in the absence of a stimulus exciting it again and again? One possibility to solve this problem is a circuiting excitation: Assume, for the sake of simplicity, a neuron represents a number you want to keep in memory for some seconds. I refer to this neuron as R (= representing neuron). R fires when the number is perceived (read or heard), and the action potential runs from R to a ‘memory neuron’ M (among others). Now, M gets an additional action potential from the prefrontal cortex (= the will to hold the number in memory). Assume, M is now sufficiently excited and fires back to R, which may, in our model, be sufficient to make R again fire to M, and so on – until the prefrontal cortex stops exciting M. This is, of course, a very simple model; in actual fact, there may be much more neurons involved in a such a circuit.
In the figure, it is not the prefrontal cortex that directly fires to the ‘memory neuron’ M, but it is attention control that excites M. Below I reason why I think that memory depends on attention – either on conscious attention, i.e., the will to hold an information in memory, or on automatic attention control, e.g., via the thalamus. For the sake of clarity, I depicted M separate from attention control, but possibly, special ‘memory neurons’ do not exist, and short-term memory is simply a function of attention control (read more).
Why do I think that working memory depends on (or is a function of) attention control? We know from experience that if we want to hold a telephone number in memory in order to dial it few seconds later, we must first focus on the number and attentively read or hear it to impress it in memory. If I have done so, I am able to correctly dial the number some seconds later. However, if my attention is distracted before I can dial the number – e.g., somebody comes in, asks me something, and I answer – then I have forgotten the number and must memorize it once again. Obviously, there is a relationship: Short-term memory seems to be depending on attention to the information that shall be memorized.
However, not only the will to hold an information in memory can activate working memory. As depicted in the figure, a perception itself also can trigger this function via an unconscious valuing system – either by its own intensity (e.g., loud music, dazzling colors) or because the valuing system deems the perception important (e.g., if it is associated with danger, food, or sex). In both cases, the unconscious valuing system focuses attention to this perception even if it is physically of low intensity. A further possibility is that, in a behavioral sequence, the sensory feedback of the last executed step(s) of the sequence is kept in working memory as an information basis for the next steps, for example, in musical improvisation and in speaking. In such cases, the behavioral sequence includes an automatic memory function, since the unconscious valuing system has learned to deem this information important and to make attention control hold it in memory.
After having considered the relationship between working memory and attention, we now can ask: What is the role of the fiber tracts between the brain regions of speech perception and speech production that were topic in the last section (Fig. 14) and will be topic in the next sections? If we assume that these tracts enable to involve sensory feedback in speech planning and -control (sometimes referred to as sensorimotor integration), then this cannot mean a data transfer (see above). Therefore, I think the fiber tracts serve for attention and working memory – functions that are basic for the control of sensorimotor sequences. At least, there is some evidence that the brain areas connected by the the two fiber tracts are active in specch processing tasks involving verbal working momory (Fengler, Meyer, & Friederici, 2015; 2016; Meyer et al., 2015) (read more).
Further, there is some evidence for a relationship between white matter develpment and attention / working memory in general. Nagy, Westerberg, & Klingberg (2004) found that the development of working memory capacity in healthy children aged between 8 and 18 years was positively correlated with fractional anisotropy in two regions in the left frontal lobe, including a region between the superior frontal and parietal cortices (see also Klingberg, 2006 for a review). Short et al. (2013) investigated 12-month-old infants and found better visuospatial working memory scores to be significantly associated with higher fractional anisotropy and lower radial diffusivity (suggesting higher myelination) in several white matter tracts, among them corpus callosum, arcuate fasciculus, and the temporal-parietal segment.
Charlton et al. (2010) investigated the relationship between working memory and white matter in aging. They identified significant clusters in the white matter that were associated with working memory performance. The tracts that passed through these clusters included the superior parietal lobule pathway and the medial temporo-frontal pathway, among others. Finally, as already mentioned in Section 4.1, Takeuchi et al. (2010) showed that the amount of working memory training of young adults correlated with increased fractional anisotropy in white matter regions adjacent to the intraparietal sulcus and the anterior part of the body of the corpus callosum after training, i.e., in regions that are thought to be critical in working memory.
Wu et al. (2016) investigated white matter microstructure in children with ADHD (aged 8-15 years). Compared with healthy controls, children with ADHD showed decreased fractional anisotropy and increased radial diffusivity (suggesting reduced myelination) in widespread, overlapping brain regions, mainly in corpus callosum and major tracts in the left hemisphere. Particularly, lower fractional anisotropy was associated with inhibition performance in the participants with ADHD. Interestingly, reduced inhition control was also found in behavioral studies with children and adults who stutter (see Section 3.3). I think all these results provide some evidence for the possibility that also the fibers in which fractional anisotropy is reduced in stutterers are involved in attention control and working memory.
Kaganovich, Hampton Wray, and Weber-Fox (2010) wrote:
“The strongest evidence for the involvement of working memory and attention in the disorder of stuttering comes from studies that employ dual tasks to examine the effect of cognitive load on speech dysfluency. Such studies require that their participants not only successfully divide attention between two demanding tasks, but also effectively maintain task-relevant information in working memory. For example, in a series of experiments Bosshardt (1999, 2002; see also 2006 for a review) tested the influence of concurrent linguistic and non-linguistic tasks on fluency of word repetitions and sentence production in adults who stutter and normal fluent speakers. He found that adding mental calculation to the primary task of word repetition significantly increased the stuttering rate only in some adults who stutter (Bosshardt (1999), while adding another linguistic task, such as silent reading and word memorization, affected stuttering frequency in adults who stutter more profoundly, resulting in a significant group difference between adults who stutter and normal fluent speakers (Bosshardt, 2002).”
In their own study, Kaganovich, Hampton Wray, and Weber-Fox (2010) investigated non-linguistic auditory processing and working memory update with event-related potentials (ERPs) in young stuttering children (4–6 years of age) and normal fluent controls. They found that an overall larger proportion of the stuttering children failed to show greater positivity (P3 component of the ERP) to deviant sounds. P3 is assumed to reflect an update in the current schema of the environment (or a working memory update) in response to a detected change (see, e.g., Polich 2007). Hence, working memory updating was seemingly less intensive in young stuttering children – however, the stuttering and non-stuttering groups did not differ in behavioral measures of working memory which were made prior to the EEG.
A possible explanation is that the oddball paradigm was presented as a game in order to maintain children’s attention, that is, some game cues, e.g., pictures on a monitor, were presented in addition to the pure tones of the paradigm. Now, the authors assume that children were possibly expecting these game cues while also monitoring the tones, thus it was a kind of dual task situation, and the stuttering children had greater difficulty distributing attention between the two tasks or were more distracted by the game elements. That would account for the attenuated P3 in the stuttering group, since P3 is attention-dependent.
In Section 2.3., I have assumed that auditory feedback is insufficiently processed by the brain when the speaker’s attention is too much turned away from the auditory channel, and in Section 4.4, I will assume that this insufficient processing may consist in an incomplete working memory update – the phoneme sequences of self-produced and (via auditory feedback) perceived speech units are not completely kept in memory for monitoring (comparison with the expectation of the correct sequence), which results in invalid error signals and, consequently, interruptions of speech flow.
In a recently published study, Kreidler, Hampton Wray, Usler, and Weber (2017) compared the processing of semantic violations by means of ERPs in children who had recovered from stuttering, in such who persisted, and in a control group. They found differences in the distribution of N400 component over the scalp and in the amplitude of the late positive component, suggesting “that future recovery from stuttering may be associated with earlier maturation of semantic processes in the preschool years.”
This result is interesting in our context because the detection of semantic violations depends on predictions generated on the basis of what was heard before, i.e., the preceding words of the sentence. Those words must be well perceived and kept in working memory for the comprehension of the entire sentence, but also for the detection of errors. Therefore, a delay in the maturation of semantic processing may point to a deficit in auditory processing and/or in working memory function. The poorer performance in nonword repetition observed in stuttering children (Anderson, Wagovich, & Hall, 2006; Anderson & Wagovich,2010; Byrd, Vallely, Anderson, & Sussman, 2012; Pelczarski & Yaruss, 2016; Sasisekaran & Byrd, 2013; Spencer & Weber-Fox, 2014)
points in the same direction.
A review of numerous studies by Smith and Jonides (1999) showed areas of activation during working memory tasks scattered over a large part of the whole cortex. Verbal working memory tendentially recruits more left-hemisphere areas, particularly the left perisylvian areas associated with speech perception and the phonological lexicon, but usually not prefrontal cortex (Postle, 2006). Prefrontal cortex seems to be involved in tasks requiring participants to deliberately hold information in memory, but not in tasks in which automatic working memory is measured (see also next footnote).
Postle, (2006) wrote; “The view that will be advanced here is that working memory functions are produced when attention is directed to systems that have evolved to accomplish sensory-, representation-, or action-related functions. From this perspective, working memory may simply be a property that emerges from a nervous system that is capable of representing many different kinds of information, and that is endowed with flexibly deployable attention.”
And about the role of prefrontal cortex (PFC):
“One hypothesized contribution of the PFC to working memory function, therefore, is to control the gain of activity in sensory processing areas of posterior cortex, in a manner that would minimize the disruption of working memory storage processes by suppressing the sensory processing of potentially distracting information in the environment. […] From the perspective of cognitive psychology, Engle and Kane and their colleagues propose that ‘executive attention’ is the mental construct that underlies [...] working memory performance, and they suggest that executive attention underlies much of the working memory-related activity of the PFC, particularly in situations in which interference must be overcome. From neuroscience, Passingham and colleagues argue that PFC activity during working memory tasks reflects attentional selection.” (Postle, 2006; section: PFC contributions to working memory; see there for references)
“...verbal working memory-related performance is positively correlated with GMP in the left parietal operculum extending into the posterior superior temporal gyrus.” (Abstract)
“ The MTG and neighboring parts of the ITG are proposed to be part of a brain network that stores lexical-semantic information, since these regions become active while processing semantic sentence ambiguity and during semantic working memory tasks.” (Section 5.2)
“...the structural disintegration of the posterior STG after a stroke predicts […] decreases in auditory short-term memory capacity […]... inferior parietal regions and the posterior STG show increased activation while processing sentences with increased verbal working memory load, which suggests that this area serves as a phonological buffer storing verbal material.” (Section 5.3)