The purpose of the current study is to investigate the processing of subject- and object-extracted relative clauses, henceforth referred to as subject and object relatives (see
Table 1 for examples). Past research has identified that object relatives are consistently more difficult than subject relatives (e.g., [
1,
2,
3]). We are interested in examining how individuals with dyslexia process these kinds of sentences because research into sentence processing in dyslexia is extremely limited, and thus, the first goal of the study is to determine whether individuals with dyslexia have difficulties with this particular type of syntactic construction. The second goal of the study is to contribute to the theoretical debate concerning the source of processing difficulty between subject and object relatives. Theoretical debates have identified two key issues: the first is violation of predictive expectations, which have been computationally assessed via Surprisal [
4,
5], and is very closely related to linguistic prediction (for reviews see [
6,
7]). The second source of difficulty is working memory. With object relatives, the object noun phrase must be held in memory until the reader encounters the relative clause verb, with which it is associated [
1,
3,
8,
9,
10,
11,
12,
13]. Thus, resolving the long-distance dependency is expected to incur substantial demand on cognitive resources, especially in terms of working memory. Dyslexia presents a very interesting test of these theoretical debates, because dyslexia has been associated with deficits in both working memory [
14,
15] and linguistic prediction [
16]. Thus, there is good reason to suspect that individuals with dyslexia will show both online processing and offline comprehension deficits with object-relative sentences.
In the remainder of the Introduction, we first cover the literature on dyslexia with a particular focus on sentence comprehension in dyslexia and what is known about the eye movement behaviour of individuals with dyslexia when they read. We then turn our attention to the theoretical psycholinguistics literature, and the two broad classes of processing models (memory-based and expectation-based) that make predictions about the processing difficulty associated with subject- and object-relative sentences. Finally, we present the rationale and hypotheses of the current study.
1.1. Sentence Processing in Dyslexia
Dyslexia is a specific learning disability that has a neurobiological origin and is primarily characterised by deficits in phonological skills. These deficits manifest as difficulties in single-word decoding and spelling, as well as in reading accuracy and fluency issues [
17,
18]. Phonological skill deficits affect an individual’s ability to manipulate, store and retrieve the phonemic and graphemic codes of language [
19]. Studies on dyslexia have reported syntactic issues in both oral and written language across the lifespan [
20,
21]. Impairments in the comprehension and production of complex syntax may originate from several sources. These range from broad weakness in language processing [
22] to more specific linguistic deficits, such as, phonological skills and/or semantics. Other studies have also suggested that deficits in dyslexia may arise from more basic cognitive abilities/executive functions, such as working memory [
23,
24]. Finally, it is important to keep in mind that many individuals with dyslexia do not read as much as typically developed individuals, and so, deficits in dyslexia may also be a secondary result of reduced reading experience [
25].
The current literature on sentence processing in dyslexia is extremely limited. This is important because we do not know whether dyslexic readers show difficulty in sentence processing and sentence comprehension, over and above single-word decoding difficulties [
26,
27]. There are considerable differences between reading single words and reading sentences. Comprehending sentences requires the ability to combine words together into meaningful phrases and extract compositional meaning, and is therefore, considerably different and more complex than single-word reading.
There have been several studies that have examined the eye movements of individuals with dyslexia, from investigating the basis of Pavlidis’ [
28] theory that atypical eye movements are the cause of dyslexia to the association between oculomotor control, visuo-spatial deficits [
29,
30,
31] and differences in saccadic eye movements [
32,
33]. Further studies on eye movements of individuals with dyslexia reading single words and non-words [
34,
35], sentences [
36,
37,
38] and texts [
26,
27] have shown that dyslexic readers tend to make longer fixations, shorter saccades and a greater proportion of regressive eye movements compared to non-dyslexic readers.
As mentioned previously, individuals with dyslexia show deficits in several areas, which fall along a continuum and are assumed to be linked to their problems with reading. In the current study, we focused on two key individual differences variables, which were assessed along with sentence comprehension and eye movements. The first was working memory [
14] and the second was verbal intelligence [
39,
40,
41]. We assumed that these two individual difference variables would play a role in the processing and comprehension of sentences with object-relative clauses. In order to read and understand a sentence, people need to be able to store and process information at the same time, as it requires them to combine prior information provided in the sentence to make inferences and resolve long-distance dependencies [
42]. Working memory has been suggested as a key factor in the successful comprehension of object-relative clauses [
9], and individuals with dyslexia often have deficits in working memory [
14,
15].
With respect to verbal intelligence, reading requires a broad vocabulary in order to quickly extract the correct meaning of words, and in turn, the meaning of sentences. According to Perfetti [
43], low-quality lexical representations lead to comprehension difficulty because the lack of automatic and/or precise associations, either at the junction of orthography-phonology or phonology-semantics, which causes information necessary for integrating a word into its sentential context to be unavailable at the time when it is needed. Van Dyke et al. [
40] reported that comprehension of subject and object relatives was much more related to verbal intelligence than to working memory [
39]. The same may also be true for individuals with dyslexia, who are often reported to have lower verbal intelligence [
25,
41]. In summary, we expected individuals with dyslexia to show differences both in terms of comprehension and eye movements, and thus, the first goal of the study is to test whether these predictions hold for subject and object relatives.
1.2. Psycholinguistic Theories–Relative Clauses
Several studies have established that sentences containing object relatives are more difficult to comprehend than sentences containing subject relatives [
1,
3,
44]. The difficulty can be manipulated by several factors, such as animacy and semantic similarity of the noun phrases occurring in the sentence [
1,
8,
45,
46,
47], as well as by the fact that object relatives are much less common than subject relatives [
48]. According to Gibson’s [
9] Syntactic Prediction Locality Theory (SPLT), which emphasises memory processes, it is predicted that while processing a sentence with a relative clause, more difficulty should arise at the relative clause verb (e.g.,
passed in a sentence like ‘The fisherman that the hiker passed carried the heavy gear’) [
5,
10]. On the other hand, a probabilistic expectation-based account (e.g., [
4]), which focuses on experience- and frequency-based expectations, predict earlier difficulty at the relative clause noun (e.g.,
hiker in the previous example). These differential predictions are important for two reasons. The first is that the source of the processing difficulty is distinct. One class of theory assumes working memory demands are the key factor, while the other assumes that difficulty arises from a violation of predictive expectation. The second reason is that the theories make different predictions about where processing difficulty should be incurred.
Eye movement studies on object and subject relatives have reported an increased number of regressions and longer reading times for object relatives compared to subject relatives [
3,
47,
49]. Expanding on previous eye-tracking studies, Staub [
44] reported, in a study that more closely resembled normal reading, that sentences with object relatives took longer to read than sentences with subject relatives. In particular, he showed elevated reading times at the relative verb and increased regressions from the relative noun. Based on this pattern, Staub concluded that both ‘classes’ of theories were partially correct (i.e., difficulty at the noun was in the form of increased regression, consistent with violation of expectation, and difficulty at the verb was in the form of elevated reading times, consistent with memory retrieval once the verb was encountered).
To date, there has only been one study to examine the comprehension of subject and object relatives in dyslexia. Wiseheart, Altmann, Park and Lombardino [
50] examined subject and object relatives in adults with and without dyslexia. Participants were shown a sentence and two images side-by-side on a computer screen, and they were asked to select the image that corresponded to the sentence. Wiseheart et al. [
50] showed that dyslexic readers had poorer comprehension accuracy compared to the control group. Controls were 93% accurate on subject relatives and 97% on object relatives, while dyslexics were 84% accurate on subject relatives and 84% accurate on object relatives. Note that the pattern for the object relatives in controls was in the opposite direction of what is most commonly reported in the psycholinguistics literature. Wiseheart et al. [
50] argued that dyslexics showed poorer comprehension accuracy compared to controls, as subject and object relatives place high demands on working memory and the individuals with dyslexia, in their sample, had lower working memory than did controls. This was further confirmed in an analysis in which working memory was covaried, as the effect of group was no longer significant. A key missing component in the Wiseheart et al. study was online-processing measures. Thus, we do not know whether/where dyslexic participants experienced online-processing difficulty, in addition to the offline comprehension impairments.
1.3. Current Study
As mentioned in the opening paragraph, the main goals of the current study are (1) to investigate whether individuals with dyslexia have difficulty processing and comprehending subject and object relatives, and (2) to contribute to theoretical debates concerning both the source of processing difficulty associated with object relatives, and also, the location of processing difficulty. To investigate these goals, we monitored eye movements as participants read subject and object relatives, and we administered additional tasks to determine how individual differences in working memory [
14] and verbal intelligence [
41,
51] were related to both online processing and offline comprehension.
Analyses focused on whether there were differences in the eye movement measures between participants with dyslexia and controls, and whether there were effects of verbal intelligence and working memory on comprehension and reading times. We expected participants with dyslexia to show poorer comprehension compared to controls, as well as to show differential eye movement patterns. More specifically, we expected to see longer reading times, more regressions and longer regression path durations in dyslexic participants in the key regions of the relative clause. Regarding the theoretical psycholinguistic debate, Gibson’s [
9] SPLT predicts difficulty at the verb in an object relative, as there is a ‘storage cost’ that slows processing while the long-distance dependency is unresolved. In contrast, expectation-based theories (e.g., [
4,
8]) predict difficulty at the relative noun. Thus, we focused our eye movement analyses on the relative verb and relative noun in the relative clause [
3]. If we find more processing difficulty at either the noun or the verb, then this would provide support for the theory that predicts difficulty at each location. Moreover, because we assessed individual differences in verbal intelligence and working memory, we were in a position to provide additional confirmatory evidence to support the underlying factors responsible for the processing difficulty associated with object relatives.