The Phonological and Executive Working Memory Deficit Hypothesis
In Chapter 2 I have presented and discussed the hypotheses proposed by McLoughlin (1994; 2002) and Fiorin (2010), which suggest that developmental dyslexia is related to a Working Memory Deficit.
As discussed in Chapter 3, the administration of the Working Memory Test Battery for Children, developed by Pickering and Gathercole precisely with the purpose of assessing children’s WM, has provided further interesting results in favor of the hypothesis that dyslexics suffer from WM deficits. In particular, dyslexic children have been found remarkably impaired in comparison to their peers in all the measures assessing the functioning of the Phonological Loop and the Central Executive. Instead, no significant differences have been found between dyslexics and controls in the measures assessing their Visuo-Spatial Sketchpad.
Taking into consideration these experimental data, I will present in this chapter an original proposal that is able to account for all the manifestations of dyslexia reviewed in Chapter 1, pointing to a disruption of dyslexics’ verbal and executive Working Memory skills. Implementing Fiorin’s hypothesis, indeed I will argue that dyslexics suffer from both a phonological memory impairment and an inefficiency affecting their executive functions, which leads to a disorder in their ability to carry out complex tasks that are particularly demanding in terms of processing resources.
However, before presenting the hypothesis, I would like to set Working Memory in a broader perspective, with the intent of showing that it represents a fundamental ingredient of human cognition. Specifically, ← 141 | 142 → I will argue that WM plays a crucial role in both reasoning and language comprehension, two aspects that are essential for our discussion on developmental dyslexia.
I will then present the Capacity Constrained Comprehension Theory, which has been developed by Just and Carpenter (1992; 2002) to account for linguistic deficits in individuals whose Working Memory skills are particularly poor. After dealing with these matters, I will present my proposal, that I call the Phonological and Executive Working Memory Deficit Hypothesis, and I will demonstrate how it is able to account for the deficits discussed in this book.
2. Working Memory and Human Cognition
It is now generally acknowledged that Working Memory has a fundamental importance in human cognition and that it is crucially involved in the execution of all complex activities, ranging from thinking and reasoning, to problem solving and, much interestingly for our discussion, to language comprehension.
Although Baddeley and Hitch’s (1974) and Baddeley’s (2000) Working Memory models have been criticized for being too simplistic and not readily falsifiable, their proposals offer a good starting point for the conceptualization of WM, providing evidence for a dissociation between the visuo-spatial and the verbal components and for the existence of a system capable of controlling and supervising operations.
Beyond its short-term storage capacity, in fact, it is widely recognized that WM is equipped with a pool of operational resources that executes the actual computations required, temporarily maintaining their intermediate and final products (Just and Carpenter 1992, 2002). The efficiency of an individual’s WM system determines therefore her ability to perform complex operations, as demonstrated by a study conducted by Kyllonen and Christal (1990) that examined the correlation between WM abilities and reasoning, measured with tests of fluid intelligence. The authors tested 2144 adults and they found that WM skills were highly ← 142 | 143 → correlated to reasoning as well as to processing speed1, demonstrating therefore that WM skills constitute the core of human cognition.
In their comprehensive review of the literature concerning the relationship between WM, fluid intelligence and processing speed, Fry and Hale (2000) noticed that these three components are strongly correlated in children as well. Specifically, they observed that WM, fluid intelligence and processing speed tend to develop in concert following a similar course and that children’s processing speed increases with age. As they grow up, indeed, children become able to process information more quickly, while adult levels are reached in middle adolescence. Interestingly, this developmental trend is very similar to that reported for WM skills and reviewed in Chapter 2.
An explanation for the correlation between WM and reasoning skills is provided by Just and Carpenter (2002), who propose an account based on the notion of capacity, arguing that the greater is the capacity of an individual’s WM and the more information she can handle simultaneously for problem solving. Importantly, they identify an individual’s WM with the functioning of her Central Executive, leaving aside the two modality-specific subsystems, namely the Phonological Loop and the Visuo-Spatial Sketchpad, since they are supposed to serve only for storage. As we will discuss in the following section, this assumption is also shared by Caplan and Waters, who identify the Central Executive as the “workhouse and mastermind of human cognition” (Caplan and Waters 1999: 77).
As mentioned above, a key concept in Just and Carpenter proposal is that of capacity, intended as the maximum amount of activation available in WM to support both storage and processing functions. Very briefly, an item is said to be activated when it crosses a minimum threshold value, becoming available in WM, and it can be used to perform the requested computations. An individual able to maintain higher levels of activations can then rely on greater processing resources to perform tasks, and consequently she will show greater reasoning abilities and speed. Conversely, an individual with a lower WM capacity can rely on lower activation resources and therefore she is forced to reallocate the activated elements, resulting in a worse and slower performance. ← 143 | 144 →
Similarly, when the task’s demands are particularly high and exceed the general capacity of the system, “some of the activation that is maintaining old elements will be deallocated, producing a kind of forgetting by displacement” (Just and Carpenter 2002: 132). In other words, when an individual has to perform a very difficult operation or a complex reasoning, both storage and computations may be degraded, with the consequence that the processing slows down and that some partial results may be forgotten, giving rise to errors.
In Just and Carpenter framework, then, individual differences do not depend on the architecture of the system, but rather on its capacity: a person whose WM capacity is higher will be more skilled in making inferences and in problem solving, thanks to her ability to maintain in WM all the necessary representations at once. Importantly, a significant consequence of this proposal is that individual differences arise when subjects are presented with processing demanding tasks. It is very likely, indeed, that all subjects perform similarly in easy tasks.
Another important proposal put forward by Just and Carpenter is that different processing domains rely on different pools of activation resources whose capacities are not necessarily correlated with each other: it follows, for instance, that an individual with a good verbal competence and high comprehension skills is not necessarily equally skilled in visuo-spatial complex tasks. Notice that this specification permits to explain why each individual can present strengths in one domain and weaknesses in another domain, and why patients affected by neurodevelopmental disorders can exhibit impaired processing at certain levels and normal performance in other cognitive areas.
To summarize so far, in Just and Carpenter’s account Working Memory skills, which coincide with the Central Executive functions, are supposed to be intimately linked to reasoning abilities and processing speed. However, individual differences concerning performance in distinct processing domains are also predicted, since specific cognitive abilities rely on specific and independent pools of resources. Note that their Capacity Constrained Comprehension Theory gives rise to precise predictions: if an individual’s verbal WM is particularly poor, in fact, she is expected to show a slower and more impaired performance in those tasks which are particularly demanding. Conversely, a subject ← 144 | 145 → exhibiting a greater verbal WM capacity is predicted to perform faster and more efficiently. These predictions have been tested by Just and Carpenter, whose theory has been indeed applied successfully to language comprehension, providing interesting results that will be exposed in the following section.
3. Working Memory and Language Comprehension
Just and Carpenter (1992; 2002) discussed specifically the role of Working Memory in language comprehension, presenting a theory, the Capacity Constrained Comprehension Theory, based on the concepts of capacity and activation introduced above. In their proposal, the information generated by computations or retrieved from long-term memory, is activated in WM during the comprehension process. Once the relevant information is activated (i.e. once it surpasses a certain threshold), it becomes available in WM and it can be used to perform the requested computations.
However, processing resources are needed to maintain the activation level: if the resources available are lower than the amount required to perform the task, then some of the information may be forgotten. As a consequence, the individual might need to process the sentence another time, causing a slowdown of the processing. Or, again, she might misinterpret the utterance.
Difficulties can arise, in particular, when the number of processes required to understand a sentence exceeds the general capacity of the system. Just and Carpenter emphasize that many of the processes required for the comprehension of an utterance occur in parallel, that is, they are executed simultaneously, generating partial products which are to be maintained and further manipulated in order to obtain the final meaning. Storage demands in WM are generally minimized by means of the tendency to interpret each new phrase as soon as possible when it is entered in the computation. ← 145 | 146 →
To better understand this proposal, let us analyze first a simple sentence, as that reported in (1), and then a more complex utterance, as the relative sentences in (2) and (3).2
(1) The girl caressed the cat.
In this case, the processor identifies the girl as the grammatical subject of the sentence and therefore it generates the expectation that a verb will occur. This expectation is satisfied as the verb caressed is encountered: at this point, the parser can establish the subject-verb dependency between the girl and caressed. Once this relation is established, it is not necessary to maintain the subject activated in memory any longer. Moreover, the verb caressed generates the expectation that an object will soon occur. Again, the expectation is satisfied by the presence of the object the cat. While syntactic dependencies are being computed, the human processor can further calculate other semantic and pragmatic features in order to assign the appropriate meaning to the sentence.
The processing is arguably more loaded when the subject has to compute a sentence like (2) and even more when it is presented with a sentence like (3).
(2) The boy scared the cat that the girl caressed.
(3) The boy that the cat that the girl caressed scared laughed.
As discussed in Chapter 1, the processing of relative sentences is remarkably demanding. The activated elements, in fact, are to be maintained in memory until the syntactic dependencies have been established. While computing a sentence like (3), in particular, most individuals are likely to experience problems assigning the appropriate thematic roles and establishing who did what to whom.
In Just and Carpenter, as well as in Gibson (1991; 1998), this happens because the amount of resources available is exceeded by the computational demands of the sentence. Specifically, the capacity of the system is not sufficient to maintain activated all the items encountered while parsing the sentence. As a consequence, the processing ← 146 | 147 → slows down and some information may be forgotten. Formally, Just and Carpenter argue that “if the activation propagation on a given cycle of production firings would exceed the activation maximum, then both the activation propagated and the activation used for maintainance are scaled back proportionally to their current use” (Just and Carpenter 2002: 133).
The main consequence of this proposal is that language comprehension is supposed to be constrained by an individual’s WM capacity. The authors suggest, in fact, that the amount of activation available to each individual varies depending on her WM capacity: a subject WM capacity, then, might determine the amount of resources that she can rely on, influencing therefore both the accuracy and the speed of her language comprehension skills.
In order to test this proposal, Just and Carpenter analyzed the performance of people with higher and lower WM skills on a range of different tasks. WM capacity was generally assessed by means of the Reading Span Task (Daneman and Carpenter 1980), in which the subject is asked to read a set of unrelated sentences and to recall, at the end of the series, the last word of each utterance in the correct order. Notice that the Reading Span Task is very similar to the Listening Span Task that I administered and discussed in Chapter 3, and that it determines the efficiency of the subjects’ Central Executive. In fact, it requires the individual to perform two tasks simultaneously, involving both storage and processing functions. Just and Carpenter administered a series of experimental protocols to high-span and low-span subjects3 to verify if there were significant differences amongst the groups, consistently with the predictions raised by their proposal. Specifically, they tested garden path sentences, ambiguous sentences and object relative clauses. Each experiment will be briefly reviewed in the following sections. ← 147 | 148 →
3.1. The comprehension of garden path and ambiguous sentences
Garden path sentences are utterances that can be easily misunderstood, because of a local ambiguity, which leads to an improper parsing. A typical garden path sentence is the one reported in (4).
(4) Fat people eat accumulates.
Individuals generally tend to analyze fat as an adjective modifying the noun people. Therefore, they process eat as the main verb of the sentence, establishing a syntactic dependency between fat people and eat, but they get stuck when they encounter the verb accumulates. They are then forced to reread the sentence, recognizing that people eat is actually a reduced relative clause and that accumulates is the main verb of the sentence. In other words, they have been “led down the garden path” due to the local ambiguity of fat, which can be interpreted as a noun or as an adjective. The ambiguity, in fact, disappears in (5):
(5) The fat that people eat accumulates.
Just and Carpenter tested a specific type of garden path sentences, following the proposal by Ferreira and Clifton (1986) who ideated a task in which readers had the possibility to avoid being led down the garden path resorting to nonsyntactic information. Consider (6):
(6) The defendant examined by the lawyer shocked the jury.
Sentence (6) is temporarily ambiguous, due to local ambiguity of the verb examined, which can both be the main verb of the sentence and the verb of a reduced relative clause. However, since the defendant is a plausible agent of the verb examined, the parser is more likely to interpret it as the main verb of the sentence. However, the interpretation of the verb can change radically if the defendant is replaced with an inanimate object, as the evidence in (7):
(7) The evidence examined by the lawyer shocked the jury.
While parsing (7), the reader can resort to nonsyntactic information to avoid being led down the garden path, observing that the evidence is an inanimate object and that therefore it cannot be interpreted as the agent ← 148 | 149 → of the verb examined. If the subject makes this reasoning, than she will interpret examined as the verb of a reduced relative clause.
Just and Carpenter examined reading times of high-span and low-span subjects asked to read and understand sentences like (6) and (7). Interestingly, they found that only people with a high WM span are sensitive to the nonsyntactic cue provided in sentences like (7), which indicates that examined should be interpreted as the verb of a reduced relative clause, given that the evidence cannot be its subject. Specifically, only high-span readers show a faster processing when the grammatical subject was inanimate, whereas no time differences are found between (6) and (7) in low-span subjects. This finding is strongly consistent with the predictions made by Just and Carpenter’s hypothesis: only people with high WM span have an amount of available resources that is sufficient to carry on a nonsyntactic analysis contemporarily to the syntactic parsing. Subjects with a lower span, instead, lack the resources to execute both analyses simultaneously and this prevents them from using the nonsyntactic cue to parse the sentence and to avoid being led down the garden path.
Similar results have been found with ambiguous sentences like the one reported below.
(8) The experienced soldiers warned about the dangers before the midnight raid.
(9) The experienced soldiers spoke about the dangers before the midnight raid.
The verb warned in (8) is locally ambiguous, since it is interpreted both as the main verb of the sentence and as the verb of a reduced relative clause.
However, Just and Carpenter found that only high-span subjects asked to read the target sentences were sensitive to this ambiguity, as demonstrated by their reading times, which were longer than that shown by low-span subjects. According to the authors, the slowing down of their processing is due to the fact that they are maintaining in memory both possible representations of the ambiguous sentence. The low-span readers, instead, show faster reading times while reading sentences like (8), whereas the differences between the two groups of subjects ← 149 | 150 → disappears with unambiguous sentences. This demonstrates that low-span readers represent just one, the most likely one, of the alternatives, which turned then to be the correct one.
Again, these different tendencies are consistent with the hypothesis proposed by Just and Carpenter: people with a lower WM capacity do not have enough resources to maintain both representations, and tend to immediately abandon the less likely one. On the contrary, individuals with a greater WM capacity can rely on an amount of resources that is sufficient to satisfy the additional demand of maintaining simultaneously two representations instead of only one.
3.2. The comprehension of object relative clauses
We have already observed that the comprehension of object relative clauses is particularly demanding, since it forces the subject to maintain the elements activated in memory longer in comparison to subject relative clauses (cf. Chapter 1). Just and Carpenter measured and confronted the reading times showed by high-span subjects and low-span subjects asked to read sentences like the ones reported in (10) and (11).
(10) The reporter that attacked the senator admitted the error.
(11) The reporter that the senator attacked admitted the error.
Arguably, (10) is easier to understand, since while analyzing (11) the subject has to maintain both the reporter and the senator activated in memory, demanding thus more processing resources.
The authors found that low-span subjects show longer reading times in comparison to high-span subjects and that they display even a less accurate performance. When asked to answer comprehension questions about these sentences, in fact, they respond correctly only in 64% of the cases, versus the 85% of the high-span subjects.
Summarizing, despite they spend more time processing the sentence, low-span readers are less accurate and commit more errors than high-span subjects. Again, then, the finding is consistent with Just and Carpenter’s proposal. ← 150 | 151 →
Another way to test the Capacity Constrained Comprehension Theory proposed by Just and Carpenter consists in manipulating the experimental context, adding to the comprehension task the request to maintain contemporarily an extrinsic memory load.
Specifically, subjects were asked to retain a sequence of digits or words during sentence comprehension. In their proposal, this additional verbal load is supposed to impose an extra burden on the individual’s WM, since further costs are needed to maintain it in memory while computing the sentence meaning.
King and Just (1991) found that the additional task of remembering a series of words affects significantly the comprehension of subject and object relative clauses. Specifically, the general accuracy decreases drastically as the request to perform this additional task is introduced, suggesting that both tasks compete for the same WM resources and that this reduces the total capacity of the system.
Finally, the analysis of distance effects has provided interesting results, too. The concept of distance refers to the gap occurring between two pieces of information that are to be related. A crucial part of language comprehension is that of establishing relationships between elements in the discourse. Clearly, it is more difficult to correlate two constituents that are distant from each other, since in this case the individual is forced to maintain the first item in WM for a longer period, increasing then the probability of forgetting and resulting in a lower accuracy. Establishing a relation between two pieces of information is indeed possible only if earlier material is still activated in memory and available for further computations.
Consistently, Daneman and Carpenter (1980) found that high-span subjects are able to maintain information activated longer in comparison to low-span subjects and that their greater capacity allows them to retrieve more accurately the appropriate antecedent of a pronoun. Examining these results, Just and Carpenter observe that WM capacity is crucially involved also in the construction of a coherent discourse interpretation. ← 151 | 152 →
To summarize, the experiments conducted on high-span and low-span subjects testing the interpretation of garden path and ambiguous sentences, the comprehension of relative clauses, and the effects of adding an extrinsic memory load and of increasing the distance between two elements to be related, provide strong evidence in favor of the Capacity Constrained Comprehension Theory. This suggests that the WM capacity of an individual constrains severely her comprehension skills, determining her ability to generate inferences, to represent simultaneously the different readings of ambiguous sentences and to establish relationships between distinct pieces of information.
In the preceding sections we have observed that Working Memory plays a crucial role in the comprehension of language. As anticipated in Chapter 2, the importance of WM in language had been grasped already by Baddeley, even though he did not analyze language comprehension in detail. Actually he simply mentioned that the Phonological Loop seems to be involved in the acquisition of language.
Another important step in this direction has been made by Just and Carpenter, who developed the Capacity Constrained Comprehension Theory precisely to account for the involvement of WM in linguistic computation. Specifically, we have observed that they identified the Central Executive as the component crucially involved in language processing and that they formulated a proposal based on the notion of capacity. Moreover, they provided strong evidence confirming that an individual with a greater WM capacity has more resources available to perform computations and that, as a consequence, she can process language faster and more efficiently in comparison to an individual with a lower WM capacity.
These results demonstrate on the one side that if the Phonological Loop is involved in the acquisition of vocabulary, influencing also the individual’s phonological awareness, the Central Executive is the component responsible for language comprehension. The Phonological Loop, in fact, cannot account for syntactic, semantic and pragmatic ← 152 | 153 → computations, but rather it provides only a phonological analysis of the sentence to be interpreted.
As mentioned above, one important consequence of this approach is that the effects of an impaired Central Executive functioning are expected to be more evident with linguistically complex sentences. A low-span subject may actually nevertheless have an amount of resources sufficient to compute easier structures and her impairment may therefore be manifested only in more demanding tasks.
Although the importance of WM in language comprehension is not challenged, a debate has arisen concerning the question whether there is a unitary verbal WM, deputed to compute also non-linguistic but verbal material, or instead a WM specific for linguistic processing.
The first proposal, shared by Just and Carpenter (1992; 2002), Fedorenko and colleagues (2006) and Gordon and colleagues (2002), argues that both linguistic and non-linguistic but verbally-mediated processing rely on a general resource pool dedicated to verbal WM.
The second account, instead, as proposed by Caplan and Waters (1996; 1999), claims that there is a verbal WM specialized for linguistic computations and another verbal WM deputed to the computation of verbal but non-linguistic material. Specifically, they suggest that the WM system involved in interpretive processes (i.e. syntactic and semantic computations) is a separate subsystem within verbal WM. This proposal raises the specific prediction that the addition of verbal material to be remembered should not interfere with the interpretation of a sentence, given that they rely on different pools of resources. According to the authors, in fact, the request of maintaining in memory a sequence of words or digits does not involve interpretive processes and it should not affect syntactic and semantic computation. However, this prediction has been disconfirmed by the results of the experimental protocols performed by Gordon and colleagues (2002) and, more recently, by Fedorenko and colleagues (2006; 2007). Fedorenko, Gibson and Rohde (2007), in particular, developed an interesting experiment, asking subjects to process syntactically complex sentences while performing a secondary task. In the first condition, this additional task had a verbal but non-linguistic nature, involving arithmetic integration processes; specifically, the subject was required to read a set of relative clauses and, simultaneously, to ← 153 | 154 → sum a sequence of numbers. The sentence was divided into four regions and it was presented on a computer screen in four fragments (e.g. “The janitor / who frustrated the plumber / lost the key / on the street”), whereas the numbers for the addition task were presented simultaneously above each fragment (e.g. “12 / +4 / +5 / +4”). Both the sentence and the addition presented varied in complexity; the amounts of time required by the subject to read each fragment of the sentence and to perform the respective addition were recorded. In the second condition, the secondary task was a spatial-rotation task, in which the subject was shown a circle with a colored sector. In correspondence of each of the four fragments of the sentence, the subject was instructed to imagine adding each sector to the preceding ones, remembering the angle obtained by the combined sectors. For the reader’s convenience, an example of this secondary task is reported below.
Figure 4.1. An example of the experiment developed by Fedorenko, Gibson and Rohde (2007).
As the arithmetic addition task, the spatial rotation task involves the integration of an incoming element into the original representation. Although the kind of processing required is similar, the spatial rotation task does not rely on verbal WM and therefore it should not affect performance in the reading task, differently from the arithmetic integration task, which instead should interfere with reading speed. Results are indeed consistent with both predictions: a strong interaction has been found in the first condition, where subjects had to compute the additions, indicating that this task relied on the same pool of resources as the linguistic task. Conversely, a significantly weaker interaction has been observed in the second condition, where participants had to perform the spatial rotation task. On one side these findings show once again that verbal WM and visuo-spatial WM are independent from each other, whereas on the other side they provide evidence against the hypothesis by Caplan and ← 154 | 155 → Waters that linguistic computations are served by a specific and separate subsystem within verbal WM. Conversely, results support the view that both linguistic and non-linguistic but verbally mediated processes rely on verbal WM.
To summarize so far, in these sections we have observed that Working Memory, and especially the Central Executive, is crucially involved in the comprehension of language, as well as in other complex activities. In the following paragraph I will propose that a Central Executive impairment, together with a Phonological Loop impairment, can be held responsible for the manifestations of dyslexia reviewed in Chapter 1.
4. The Phonological and Executive Working Memory Deficit Hypothesis
As discussed both in Chapter 2 and in the present chapter, Working Memory is the brain system which has the function to (i) temporarily store information, (ii) manipulate this information in order to perform cognitive tasks (iii) and to temporarily maintain the outcomes of intermediate computations. A number of studies have provided evidence for the existence of a distinction between the storage of verbal material and that of visuo-spatial (and possibly kinesthetic) material, which in Baddeley’s and Hitch’s model are handled respectively by the Phonological Loop and the Visuo-Spatial Sketchpad. Experimental results seem to indicate that a similar subdivision can be found also within the Central Executive, which is the system deputed to control attention, to supervise the activities and to perform the computations necessary to carry out cognitive tasks. Specifically, the experiment by Fedorenko and colleagues reviewed above strongly indicates that the processing of verbal material relies on the same pool of WM resources, independently from its linguistic or non-linguistic nature, whereas the processing of visuo-spatial material is served by a separate pool.
Moreover, we have observed that the Capacity Constrained Comprehension Theory proposed by Just and Carpenter (1992; 2002), consistently with a huge amount of experimental results, show that people ← 155 | 156 → with a lower verbal WM capacity will likely exhibit a slower and less accurate performance in those tasks requiring a complex processing and, therefore, a greater amount of WM resources.
As observed in Chapter 1, it seems that dyslexics are remarkably impaired in comparison to their peers just in demanding and complex tasks, besides their well-known phonological impairments. For this reason, I propose that their deficits arise from a specific impairment affecting their Working Memory, and in particular, resorting to Baddeley and Hitch’s terminology, their Phonological Loop and their Central Executive. As confirmed by the experimental results reported in Chapter 3, as well as by those reported in Chapter 2, the functioning of the Phonological Loop and of the Central Executive are severely compromised in dyslexics, who underperform in comparison to controls in all measures tapping their phonological memory and executive functions. In my hypothesis, formulated below, these impairments are to be considered responsible for the deficits manifested by dyslexic individuals.
(12) The Phonological and Executive Working Memory Deficit Hypothesis
Dyslexic individuals suffer from a limitation affecting their Working Memory and hampering in particular their phonological memory and their executive functions. As a consequence, this impairment disrupts their phonological competence, as well as their performance in complex tasks that are particularly demanding in terms of Working Memory resources.
On the contrary, dyslexics can rely on a spared visuo-spatial memory, to which they can resort for the accomplishment of compensatory strategies.
The Phonological and Executive Working Memory Deficit Hypothesis yields, therefore, the very specific prediction that dyslexic individuals are likely to underperform in all those tasks which involve a fine phonological competence and which are particularly demanding in terms of processing resources. Similarly to what predicted by the Capacity Constrained Comprehension Theory, I propose that their more limited or less efficient WM compromises their performance in complex tasks that require an amount of processing resources that overcomes their actual ← 156 | 157 → capacity. As a consequence, they display a less accurate and slower performance. Conversely, dyslexics are predicted to behave normally in easier and less demanding tasks.
Based on these premises, then, the Phonological and Executive Working Memory Deficit Hypothesis is able to account for all the deficits typically associated with dyslexia and reviewed in Chapter 1, which, indeed, share the common and crucial features of either necessitating a precise phonological analysis (as phonological awareness tasks) or a complex and costly processing (as linguistic comprehension tasks) or both (as reading and spelling). In the following sections we will discuss in detail how this hypothesis can explain reading and spelling deficits, poor phonological competence, vocabulary and rapid naming deficits, grammatical problems and attention disorders.
4.1. How the hypothesis explains reading and spelling deficits
In Chapter 1 we have observed that dyslexic children suffer from marked reading impairments, as evidenced by their slow, inaccurate and effortful reading. Their error patterns typically seem to reflect a poor capacity to visually discriminate similar graphemes and a tendency to substitute similar looking words. Remarkable difficulties arise in particular when they are asked to read nonwords or very long words, whereas they perform better with highly frequent and shorter words. Moreover, we have noted that deficits appear to be more marked in those children whose mother-tongue has an opaque orthographic system like English.
Analyzing the Dual-Route Model of reading developed by Coltheart (1985), we have observed that dyslexic children appear to rely more heavily on the lexical route, which processes familiar words that are already stored in the semantic system, as evidenced by their better performance with frequent words. In the lexical route, in fact, stimuli are processed as indivisible units and their phonological form is thus accessed directly. Problems are more marked, instead, when dyslexics are forced to use the sublexical route, which breaks down the string in its minimal components, resorting to a set of orthographic-phonological ← 157 | 158 → conversion rules to retrieve the pronunciation of each unit. This route is obligatorily used to read nonwords and less familiar words, whose phonological form is not stored in the semantic system. The deficits shown by dyslexics in nonword reading tasks suggest that their sublexical route is not functioning properly, pointing to a difficulty at establishing the correct relationships between graphemes and phonemes. Similarly, we have noted that Frith’s (1986) model of learning to read predicts that the deficits exhibited by dyslexics are caused by a failure to successfully master the conversion rules that are normally acquired in the alphabetic stage, suggesting that they are not able to perform the requested phonological analysis.
In both models, then, it is predicted that dyslexics’ deficits are due to a phonological weakness preventing them from acquiring the orthographic-phonological conversion rules. This weakness can be explained by the Phonological and Executive Working Memory Deficit Hypothesis, if we consider that Working Memory is crucially involved in phonological analyses as well as in complex tasks. Reading and spelling are namely not only activities that necessitate a good phonological competence, but they are also complex tasks that require a considerable amount of processing resources.
Let us analyze, for instance, the process required to read a stimulus in a transparent language like Italian using the sublexical route. In this case, the child has to: (i) decompose the string, (ii) identify each grapheme, (iii) retrieve and apply the appropriate grapheme-phoneme conversion rule for each grapheme composing the stimulus and (iv) maintain the intermediate products activated in memory in order to blend phonemes together as long as the stimulus has been completely processed. Arguably, this process protracts longer with longer words, hence the minor difficulty experienced by dyslexics with shorter stimuli. Things can be further complicated in presence of irregular clusters, such as the Italian gn cluster in a word like sogno (‘dream’). In this case, the child cannot operate the normal grapheme-phoneme conversion rule that would assign to the grapheme “g” the phoneme /g/ and to the grapheme “n” the phoneme /n/. Rather she has to retrieve the orthographic-phonological conversion rule according to which the cluster “gn” is to be translated into a single phoneme, namely /ɲ/. Arguably, then, reading becomes even more ← 158 | 159 → demanding in presence of irregular words, as evidenced by the greater difficulties experienced by dyslexics in these cases. Consequently, it is not hard to believe that children have even more troubles at learning to read in a language like English, where there are more possible mappings between graphemes and phonemes. Conversely, the child’s task is facilitated in the presence of highly familiar words, whose phonological forms have already been stored and can be retrieved directly through the lexical route.
Once we have established that reading is indeed a complex and resource-demanding activity, it appears clear that a normally functioning Working Memory constitutes a fundamental prerequisite for the proper acquisition of the correspondences between graphemes and phonemes. Those children who suffer from WM limitations, i.e. by hypothesis dyslexic children, will arguably show more difficulties in learning to read in comparison to normally achieving children that can rely on a more efficient WM, exhibiting a slower and less accurate reading. Fortunately, these difficulties tend to disappear, or at least to diminish, as the child starts to automatize the process, relying more heavily on the lexical route and therefore bypassing the harder and more laborious processing involved by the sublexical route. However, difficulties reappear as soon as they are presented with nonwords, as indicated by the fact that even adult dyslexics, whose reading competence has achieved a satisfactory level, have troubles in reading novel or non- existing stimuli.
To summarize, then, the difficulties in learning to read experienced by dyslexics can be ascribed to an impairment affecting both their phonological memory and executive functions (i.e. Phonological Loop and Central Executive in Baddeley and Hitch’s terminology).
A similar argument can be put forward to explain spelling difficulties, which often co-occur with reading deficits in developmental dyslexia. As observed in Chapter 1, the mastery of phoneme-grapheme conversion rules is essential for the appropriate acquisition of spelling skills. Typical misspellings shown by dyslexic children concern omissions of consonants in complex clusters, omissions of double consonants, confusion of similar graphemes or phonemes and incorrect translations of irregular spellings, especially in languages whose ← 159 | 160 → orthographic system is opaque, as English. As in reading, children’s difficulties increase when they have to spell longer and less frequent words or nonwords, whereas they perform better with shorter and familiar words, suggesting that they can rely on already stored orthographic forms to retrieve the correct spelling of familiar words. An impaired phonological analysis, then, affects detrimentally the child’s spelling abilities.
Furthermore, spelling is a complex task too, requiring the subject to perform a sequence of steps, applying phonological-orthographic conversion rules and maintaining in memory the products of intermediate computations. Arguably, then, WM is crucially involved in this activity: errors and slowness increase proportionately to the processing costs required to perform the task, that is, with longer, unfamiliar and irregular words. The typical misspellings reported above, then, can be interpreted as the consequence of a cognitive overload: the elements that are more frequently forgotten, in fact, are often “redundant” items, such as double consonants or silent graphemes.
Summarizing, then, we can maintain that dyslexics’ reading and spelling deficits are caused by their limited phonological memory and executive functions, as argued by the Phonological and Executive Working Memory Deficit Hypothesis.
4.2. How the hypothesis explains phonological deficits
In Chapter 1 we have reviewed a number of studies demonstrating that phonological deficits are very widespread amongst dyslexic people. 100% of dyslexics exhibit, in fact, a very poor phonological awareness, as assessed by metalinguistic tasks. Typical phonological awareness tasks require the subject to identify the initial, final or middle sound of words, to detect and produce words that rhyme, to decompose words into syllables and sounds, to blend syllables and sounds into words and to delete or substitute syllables or sounds in words. A compelling body of evidence, indeed, has confirmed that dyslexics perform very poorly in phonological tasks and that their phonological awareness is significantly low, suggesting that they meet difficulties in analyzing the internal structure of words. ← 160 | 161 →
Moreover, poor phonological competence in developmental dyslexia persists across ages, as proved by the fact that it is displayed both by children and by adults, and across different languages. As argued, amongst others, by Ramus and Szenkovits (2008), as well as by Ramus and Ahissar (2012) and Ramus et al. (2013), the deficient phonological ability displayed by dyslexics seems to be caused by a specific deficit affecting the access to phonological representations, which are in themselves intact and fully specified. This proposal is perfectly in line with the Phonological and Executive Working Memory Deficit Hypothesis, which argues that dyslexics exhibit an impairment affecting their phonological memory. As emphasized by Fiorin (2010) in his Verbal Working Memory Deficit Hypothesis, in fact, the Phonological Loop is precisely concerned with the abstract representations of the sounds composing language and it plays a crucial role in accessing those representations.
Since it has been demonstrated that the Phonological Loop is remarkably impaired in dyslexic children, it is reasonable to propose that this impairment is responsible for the deficits they exhibit in phonological awareness tasks.
4.3. How the hypothesis explains vocabulary and naming deficits
Another deficit typically reported in dyslexic individuals concerns their vocabulary development and their performance in rapid naming tasks, in which they are required to rapidly name pictures of colors, objects or alphanumeric characters (see Chapter 1). Specifically, we have noted that dyslexic children’s vocabulary is often underdeveloped in comparison to that of age-matched controls and that they display significant length and frequency effects, remembering better short and frequently used words.
Again, this deficit can be captured by the Phonological and Executive Working Memory Deficit Hypothesis: as demonstrated by Baddeley and colleagues, in fact, WM, and in particular the Phonological Loop, has the fundamental function of supporting the long lasting acquisition of new words (see Chapter 2). Dyslexics’ poorly functioning Phonological Loop may then hamper their vocabulary learning, affecting even more ← 161 | 162 → evidently the acquisition of long and infrequent words. Arguably, this deficit affects also the learning of foreign language vocabulary, causing the difficulties often exhibited by dyslexic children in acquiring a second language.
Beyond their poor vocabulary knowledge, dyslexics have been found remarkably impaired in rapid naming tasks across ages and languages: both children and adults are namely much slower at picture naming than chronological age-matched unaffected individuals and even than reading age-matched controls. Interestingly, the greatest difficulties appear when they are asked to quickly name pictures of objects in comparison to alphanumeric characters or colors. These results suggests that the retrieval is slower in dyslexic individuals and that slowness increases proportionately to the number of possible responses.
This lexical access deficit can be viewed as consequence of a poorly functioning WM. As shown by Baddeley, WM is intimately linked to Long-Term Memory: in his model, specifically, the connection relating the Phonological Loop and language indicates that the Loop is also engaged in the retrieval of the words that are already permanently stored. The impairment affecting dyslexics’ WM could then hamper their ability to retrieve the words stored in LTM, mainly resulting in a general slowing down of the process. Moreover, the greatest difficulty reported when subject are asked to quickly name objects in comparison to alphanumeric characters and colors can be interpreted as a consequence of the greater amount of resources required. There is, in fact, a limited number of colors, digits and letters, whereas the number of possible alternatives increases radically in the case of objects, enhancing proportionately also the processing costs of the operation performed.
In Chapter 1 we have observed that dyslexics’ linguistic difficulties are not confined to the domains of phonology and vocabulary, showing that they extend to grammar as well. Specifically, we have noted that dyslexic children are remarkably impaired in comparison to controls in the interpretation of tough sentences, pronouns, relative clauses, passive ← 162 | 163 → sentences and grammatical aspect and that they exhibit morphosyntactic deficits as well. In this paragraph I will argue that these deficits appear to reflect a processing difficulty affecting dyslexic individuals, as predicted by the Phonological and Executive Working Memory Deficit Hypothesis. Specifically, I will refer to the Capacity Constrained Comprehension Theory developed by Just and Carpenter and discussed above, suggesting that individuals with a lower Working Memory Capacity4 are more likely to manifest deficits in the interpretation of those sentences which are particularly demanding in terms of processing resources. We have observed, in fact, that difficulties arise when the processing cost of the operations to be performed exceeds the general capacity of the system: if the parser is not able to process or maintain activated all the items necessary for the actual comprehension of the sentence, then the processing will slow down and some information, as the intermediate products of previous computations, may be forgotten.
According to the Phonological and Executive Working Memory Deficit Hypothesis, then, dyslexics are likely to manifest a less accurate performance or even to get stuck when asked to carry out demanding operation and to interpret linguistically complex sentences. Bearing this in mind, we can now try to provide an explanation to the grammatical deficits reported by dyslexic children, considering the results of the experiments presented in Chapter 1.
4.4.1. The Interpretation of Tough Sentences
Let us consider first the interpretation of tough sentences, reexamining the examples discussed in Chapter 1 and reported below.
(13) a. The snake is glad to bite.
b. The snake is hard to bite.
c. The snake is horrible to bite.
As we have noted, both acquisition studies and the experiments conducted by Byrne (1981) have revealed that O-type constructions, such as (13b), are more difficult to interpret for both young children and dyslexics, who show instead a normal performance with S-type sentences, ← 163 | 164 → like (13a). Moreover, both groups of children manifest the tendency to interpret utterances like (13c), which are ambiguous between the two types of interpretations, as S-type constructions, differently from typically developing children matched for chronological age with dyslexic children. Arguably, the discrepancy between the two types of constructions can be ascribed to processing factors: in (13a), in fact, the snake is both the grammatical subject of the utterance and the logical subject of the action of biting. The parser is then simply required to compute the dependency between the subject the snake and the verb to bite. In (13b), instead, the snake is the object of the action of biting. The interpretation of the sentence, then, involves the computation of two, instead of only one, syntactic dependencies: the role of the subject, in this case, is carried by a silent (i.e. not phonetically realized) pronoun, traditionally indicated, in the syntactic literature, as PRO. Therefore, the parser must compute first the dependency between PRO and the verb to bite and then the dependency between the verb and the object the snake.
Therefore, O-type constructions are more difficult to interpret than S-type constructions and more demanding in terms of processing resources. The greater difficulty experienced by dyslexics and young children, together with the higher tendency to interpret ambiguous sentences as S-type constructions, suggests then that their processing resources are more limited. This is consistent both with the Capacity Constrained Comprehension Theory, claiming that individuals with a lower WM capacity are expected to manifest deficits with complex constructions, and with the Phonological and Executive Working Memory Deficit Hypothesis, arguing that dyslexics suffer precisely from a WM limitation.
4.4.2. The interpretation of pronouns
In Chapter 1 we have observed that dyslexic children are remarkably more impaired than age-matched typically developing children in the interpretation of pronouns, performing as younger children. Waltzman and Cairns (2002) found that they are impaired in the comprehension of sentences like (14), accepting as grammatical a sentence like (14b) in a context in which Anna admires herself, in violation of Principle B of the Binding Theory. ← 164 | 165 →
(14) a. Annaj admires heri.
b. *Annaj admires herj.
Discussing this finding, we have adopted Grodzinsky and Reinhart (1993)’s proposal according to which the interpretation of sentences like (14) involves the computation of Rule I, reported below, which is very demanding in terms of processing resources.
(15) Rule I
α and β cannot be covalued in a derivation D, if
a. α c-commands β
b. α cannot bind β in D, and
c. The covaluation interpretation is undistinguishable from what would be obtained if α binds β.
[To check c, construct a comparison-representation by replacing β, with a variable bound by α].
According to Grodzinsky and Reinhart the computation of Rule I is necessary to determine if a sentence like (14b) is grammatical or not, and it is very costly in terms of memory resources. In order to apply Rule I, in fact, an individual has to perform three different steps: in this case, for instance, she has first to verify if her is c-commanded by Anna and secondly if it can be bound by Anna. Once ascertained that the pronoun is c-commanded by the DP, she has to construct a comparison-representation to establish if the two readings, obtained respectively by covaluation and binding, are equivalent or not. This third step is supposed to be the most expensive, since it requires the subject to construct, maintain in memory and compare two distinct representations. This operation, known as Reference Set Computation, is arguably very demanding in terms of WM capacity (see Chapter 5). Specifically, Grodzinsky and Reinhart argue that young children lack the resources necessary to carry out this task. Arguably, this is due to the fact that their WM is still developing.
As young children, dyslexics seem to be unable to compute Rule I, committing more errors than their peers. In the framework of the Phonological and Executive Working Memory Deficit Hypothesis, this is due to the fact that their low WM capacity prevents them from performing cognitively complex operations, like the Reference Set Computation, ← 165 | 166 → consistently with what predicted by the Capacity Constrained Comprehension Theory.
This hypothesis permits also to explain the results reported by Fiorin (2010) who tested sentences that were ambiguous between the bound-variable and the coreferential reading, as (16).
(16) Every friend of Francesco painted his bike.
Binding: For every x, if x is a friend of Francesco, then x painted x’s bike
Coreference: For every x, if x is a friend of Francesco, then x painted Francesco’s bike
Fiorin found that dyslexics, unlike control children, displayed a constant tendency to assign the same interpretation to ambiguous sentences, and he proposed that this tendency was a consequence of their limited processing resources. Dyslexics’ propensity to stick to the same interpretation can be seen as a strategy adopted to avoid the expensive process of resolving the ambiguity of the sentence by applying Rule I and, thus, deriving both possible representations.
Once again, these results can be accounted for by the Phonological and Executive Working Memory Deficit Hypothesis.
4.4.3. The Interpretation of Relative Clauses
As discussed above, the Capacity Constrained Comprehension Theory predicts low-span individuals to manifest considerable deficits in the interpretation of object relative clauses. The comprehension of this kind of sentences requires the subject to maintain two DPs (instead of only one as it happens with subjects relatives) activated in memory in order to establish the appropriate syntactic dependencies. Interestingly, it has been found that, similarly to the low-span subjects tested by Just and Carpenter, dyslexics have troubles with object relative clauses, as the one reported in (17b).
(17) a. The lion hugs the bear that rolls the ball.
b. The bear that the lion hits rolls he ball.
To compute the syntactic dependencies of (17a) the parser has to keep in memory only one DP at once: the lion is maintained until the verb ← 166 | 167 → hugs is encountered and then the bear has to be maintained until rolls is found. To interpret (17b), instead, the parser needs to maintain both the bear and the lion simultaneously activated in memory in order to establish their relations to the verbs hits and rolls. Arguably, then, the computation of (17b) is more costly: people with a lower WM capacity are then supposed to be more impaired than people with higher processing resources available to perform the operation. The fact that dyslexic children underperform in comparison to age-matched typically developing children while computing object-relative clauses is then compatible with the Phonological and Executive Working Memory Deficit Hypothesis, arguing that they lack the processing resources required to interpret this kind of syntactic structures. Conversely, they do not manifest problems with the interpretation of simpler structures like (17a), whose lower demand of resources can be successfully handled by their WM.
4.4.4. The Interpretation of Passive Sentences
In Chapter 1 we have discussed the results of the experimental protocol administered by Reggiani (2010) to test the interpretation of passive sentences in dyslexic children. Results have demonstrated that dyslexics are remarkably more impaired than age-matched controls in the interpretation of reversible non-actional passives, like that reported in (18), performing at the same level of younger preschool children.
(18) Winnie the Pooh is seen by the bees.
Specifically, sentence (18) was judged grammatical in a context in which Winnie the Pooh saw, without being seen, some bees. To explain this result, Reggiani argued that dyslexics’ poor performance is caused by a processing deficit. The interpretation of passive clauses such as (18), in fact, requires the subject to handle both the non-canonical word order typical of passive sentences and the use of a psychological non-actional verb like to see. Performing like younger children, dyslexics appear then to show the so-called Maratsos Effect, which arises precisely when these two ingredients are to be computed at once.
Along these lines, Reggiani suggests that the computation of passive sentences constructed with non-actional psychological verbs constitutes a too difficult task for both dyslexic and control children, whose ← 167 | 168 → WM capacity is not large enough to cope with this kind of operation. This would entail, again, that results can be accounted for by the Phonological and Executive Working Memory Deficit Hypothesis.
4.4.5. The Interpretation of Grammatical Aspect
The interpretation of grammatical aspect in dyslexic children has been tested by Fiorin (2010). As discussed in Chapter 1, he tested the interpretation of the Italian past tenses Imperfetto, which encodes the imperfective aspect, and Passato Prossimo, which encodes the perfective aspect, in sentences like those reported below:
(19) a. IMPF: Marco mangiava il gelato.
‘Marco ate-IMPF the ice-cream’.
b.PP: Marco ha mangiato il gelato.
‘Marco ate-PP the ice-cream’.
Dyslexics displayed a significantly poorer behaviour in comparison to age-matched normally achieving children when asked to interpret sentences encoding the imperfective aspect, accepting (19a) as a correct description of a picture that portrays Marco having already finished eating the ice-cream. Conversely, they interpreted as well as controls sentences like (19b).
To explain these results, Fiorin argued that the association of IMPF sentences with ongoing situations is very expensive, since it requires the subject to perform a complex reasoning, computing a conversational implicature. Specifically, she has to reason that if the speaker had wanted to refer to a complete situation, she would have used a PP sentence, since it would be more informative: whereas IMPF sentences can be used both for complete and ongoing situations, in fact, PP sentences can only describe complete events. Therefore, the speaker’s choice to use the less informative IMPF construction indicates that the PP would have not been appropriate and that the complete situation is to be discarded in favor of the ongoing situation. As we will discuss in Chapter 5, Reinhart (1999; 2006) showed that the computation of conversational implicatures is costly in terms of processing resources, since it involves a Reference Set Computation, requiring the subject to construct and compare the two alternative interpretations of the sentence. ← 168 | 169 →
Consistently with the Capacity Constrained Comprehension Theory, the Phonological and Executive Working Memory Deficit Hypothesis can explain dyslexics’ difficulties, arguing that they lack the WM resources necessary to accomplish this task. Their better performance with PP constructions further confirms that their difficulties arise in presence of structures that are costly in terms of the involved processing resources, whereas they perform normally with less demanding structures.
4.5. How the hypothesis accounts for morphosyntactic deficits
The presence of morphosyntactic deficits in dyslexic children has been reported in the experiments administered, amongst others, by Joanisse and colleagues (2000), Jiménez and colleagues (2004) and Rispens (2004). Results showed that dyslexics’ morphosyntactic competence is highly impaired in comparison to that displayed by both chronologically age-matched and reading age-matched normally achieving children. Specifically, problems arose when children were asked to apply morphosyntactic rules, such as those concerning the formation of past tenses and plurals.
Again, findings seem to evidence a disruption in dyslexic children’s ability to apply rules: similarly to what happened in the acquisition of the grapheme-phoneme conversion rules necessary for correctly reading and spelling, dyslexics appear to be slower and more impaired than typically developing children in acquiring and applying morphosyntactic rules.
In the Phonological and Executive Working Memory Deficit Hypothesis, this impairment can be interpreted as a consequence of dyslexics’ poorer Working Memory: the lack of an appropriate amount of cognitive resources hampers their speed and efficiency in the acquisition and application of rules. However, this does not mean that dyslexics will never be able to apply those rules, but simply that their ability will develop more slowly. As soon as the rules get automatized, in fact, the amount of processing resources required to apply them is expected to decrease noticeably. ← 169 | 170 →
4.6. How the hypothesis accounts for attention deficits
In Chapter 1 we have observed that dyslexics tend to exhibit an inattentive behaviour, characterized by lacks of concentration, high levels of distractibility and short attention spans. We have noted that attention is the ability required to direct the necessary resources to a task, focusing on the relevant information and filtering out irrelevant material. Experimental protocols specifically designed to detect the presence of attention deficits in dyslexics have revealed that they are more impaired than typically developing children in tasks testing their attention and their resistance to interference, as the Stroop Task. In this task, in particular, dyslexic children manifest remarkably higher levels of interference in comparison to controls when asked to name the ink color of words printed in a tint, which was different from that denoted by the word itself.
The attention deficits displayed by dyslexic individuals can be accounted for by the Phonological and Executive Working Memory Deficit Hypothesis, which claims that, beyond a phonological memory impairment, dyslexics suffer from a limitation affecting their executive functions. Remember that first Baddeley and Hitch (1974) and then Baddeley (1996; 2000) argue that a fundamental function of the Central Executive consists in controlling attention, an ability which plays an essential roles in all cognitive tasks, and especially in those situations in which more tasks have to be performed simultaneously.
Since dyslexics’ Central Executive functioning has been found impaired, it appears evident that the attention deficits reported represent a direct consequence of this impairment. Moreover, the inattentive behaviour frequently displayed by dyslexic children is a further effect of their poor executive functioning, which is responsible for their tendency to forget the content of instructions, to lose the focus of the task and to abandon an activity before having completed it, together with their difficulty in shifting attention from one task to another and in organizing their activity. ← 170 | 171 →
In this chapter I have presented the Phonological and Executive Working Memory Deficit Hypothesis, arguing that people affected by developmental dyslexia suffer from an impairment hampering the efficiency and the capacity of their phonological memory and executive functioning. Moreover, I have proposed that these deficiencies can be held responsible for the manifestations of dyslexia discussed in Chapter 1: specifically, I suggested that dyslexics’ difficulties are caused either by a phonological impairment, which can be traced back to an impaired Phonological Loop, or by the excessive complexity of the task. Adopting the perspective put forward by Just and Carpenter (1992; 2002) in their Capacity Constrained Comprehension Hypothesis, I argued that dyslexics’ more limited WM (or Central Executive) capacity prevents them from performing those tasks whose cognitive demands exceed their processing resources. Specifically, I suggested that they lack the cognitive resources necessary to carry out complex and demanding tasks, determining a general slowness and worsening of their performance. On the contrary, as predicted, they perform normally in those tasks for which they have a sufficient amount of resources available.
An important consequence of this hypothesis, claiming that individual abilities are determined by the individual’s Working Memory capacity, is that different degrees of severity are supposed to arise. It is not expected, in fact, that all dyslexic individuals will perform in the same manner, as their behavior is essentially dependent on their individual cognitive capacities. Dyslexics with a less severe WM limitation, for instance, are likely to manifest less marked difficulties in complex tasks in comparison to dyslexics with a more severe deficit. Arguably, then, the severity of the disorder is determined by the general damage affecting the system. This permits to explain the great differences and variability found within the dyslexic population.
Another interesting aspect of this hypothesis, thus, is that it can account for individual differences without assuming that they are determined by a difference in the underlying architecture of the system.
Furthermore, an important qualification should be made: the majority of the data discussed in this book concerns experiments that ← 171 | 172 → have been administered on dyslexic children. As we have observed in the preceding chapters, however, children’s Working Memory is not yet fully developed, since individuals’ WM is expected to increase until adolescence. This means that their performances are predicted to enhance with age, and that their deficits are likely to become more attenuated or even to disappear as their WM’s development completes. However, this does not imply that they will catch up completely with their peers in every aspect: conversely, dyslexics are supposed to keep showing a more limited WM in comparison to normally achieving individuals, although it is very likely that it will go unnoticed in everyday activities. The impairment, in fact, will be evident only when they will be engaged in relatively complex tasks whose computation exceeds the general processing capacity of the system, without interfering with the execution of simple, or relatively simple, activities.
Moreover, it is important to remember that compensation is allowed: it has been proven, indeed, that individuals can bypass their difficulties by resorting to a better functioning system to overcome the poor functioning of another system (see Chapter 2). As evidenced by Hulme and Roodenrys (1995), in fact, memory problems “may not have devastating consequences for cognitive development in the presence of adequate compensatory resources” (392). An individual with a high IQ score, thus, can be able to learn new strategies to bypass her difficulties, achieving anyway a good and satisfactory performance. An interesting example in this respect is provided by the reading strategies shown by those dyslexics who learn to read relying more heavily on the lexical route in order to bypass their impaired sublexical route.
To conclude this chapter, once ascertained that dyslexics’ difficulties arise in particular in demanding task, I decided to further test the Phonological and Executive Working Memory Deficit Hypothesis by assessing dyslexics’ linguistic ability in the comprehension of complex constructions. With this purpose, I developed and administered three experimental protocols assessing linguistic aspects that have proven to be particularly challenging in terms of processing resources, like the computation of scalar implicatures, the comprehension of negative sentences and the reference assignment to both zero and phonetically realized pronouns. The results of these experiments will be presented and discussed respectively in Chapter 5, Chapter 6 and Chapter 7.
1 Processing speed can be defined as the speed at which an individual is able to perform and complete a task.
2 Cf. with the processing of relative sentences discussed in Chapter 1. The model proposed by Gibson (1991, 1998) shares the assumptions of the approach proposed by Just and Carpenter and reviewed in this paragraph.
3 The spans of the Reading Span tests typically range from 2 to 5.5 words for sentences such as “When at last his eyes opened, there was no gleam of triumph, no shade of anger”. High span subjects have spans of four words or more, whereas low span subjects have spans of less than three words. Medium span subject, instead, have spans of three and three and a half words (Just and Carpenter 1992).
4 As Just and Carpenter do, I will use the label Working Memory (WM) to refer to Baddeley and Hitch’s Central Executive.