Chapter 6 revisits the grammatical asymmetry, a key effect in agreement attraction research. The grammatical asymmetry refers to the phenomenon where attraction effects occur in ungrammatical sentences but not in grammatical ones. This chapter evaluates existing evidence, particularly in response to challenges raised by Hammerly et al. (2019), who claimed that the empirical evidence for the asymmetry is not particularly strong and that the effect could be a product of response bias rather than an inherent property of agreement attraction. Through a detailed review of over ninety experiments, the chapter finds strong support for a grammatical asymmetry, as predicted by the retrieval-based account. Additionally, it explores how altering the ratio of ungrammatical to grammatical fillers in experiments can influence retrieval mechanisms and artificially produce a symmetrical attraction profile, yielding the response bias effect observed by Hammerly et al. These findings suggest that a symmetrical profile could emerge from increased uncertainty in memory retrieval rather than faulty linguistic representations, offering a nuanced interpretation of existing behavioral findings.

Chapter 4 introduces the leading accounts of attraction. Representational-based accounts claim that attraction arises from errors in the encoding of linguistic structures in working memory, where misrepresented features like plural number are mistakenly bound to the wrong item in memory. Feature percolation and spreading activation drive feature misrepresentation, both contributing to attraction effects. In contrast, retrieval-based accounts claim that while the sentence’s structure is faithfully encoded, attraction occurs during memory retrieval, where nontarget items that partially match retrieval cues (such as plural number) are erroneously selected for dependency formation. The chapter addresses the strengths and limitations of each approach, discussing the model predictions and the empirical evidence used to arbitrate between the accounts. Finally, it suggests that representational and retrieval-based accounts are not mutually exclusive; a comprehensive understanding of sentence processing must consider the interactions between encoding and retrieval mechanisms. This theoretical review sets the stage for evaluating empirical findings discussed in subsequent chapters.

Chapter 1 introduces linguistic illusions, focusing on how the mind processes language in real time and how systematic errors, such as agreement attraction, occur. The chapter first explains how linguistic illusions are cases where listeners or readers misunderstand or fail to notice anomalies in language. Agreement attraction, a phenomenon where mismatched subject–verb agreement is overlooked due to interference from nearby elements, serves as the primary case study. The chapter draws parallels between linguistic illusions and optical illusions, emphasizing that while both reveal discrepancies between perception and reality, linguistic illusions are more probabilistic and context dependent. This chapter also sets up the importance of studying these illusions to uncover fundamental cognitive mechanisms and processes underlying language comprehension. By systematically analyzing linguistic illusions, researchers can gain deeper insights into the cognitive architecture of language and the role of memory encoding and retrieval in language processing. The chapter concludes by outlining the book’s structure and key questions that the study of linguistic illusions aims to answer.

Chapter 5 evaluates the leading theories of agreement attraction by comparing their ability to explain key empirical findings. The chapter examines four major effects: the markedness asymmetry, grammatical asymmetry, timing asymmetry, and attraction beyond number agreement dependencies. Through detailed comparisons, the chapter highlights how retrieval-based accounts provide the broadest empirical coverage, successfully explaining each effect, while representational-based accounts mainly capture the markedness asymmetry. The chapter also introduces evidence from studies on semantic and morphosyntactic attraction, showing that retrieval-based models offer a more unified explanation of these effects across linguistic domains. Additionally, the chapter discusses evidence of number misinterpretation, which is uniquely predicted by representational accounts, but suggests that these effects may be task-specific artifacts of metalinguistic processes. This theoretical arbitration provides a comprehensive overview of the strengths and limitations of both accounts and emphasizes the need for further research to fully understand the cognitive mechanisms underlying attraction phenomena.

In this chapter I present the main elements of the theory of bare phrase structure: principally the basic operation Merge. This operation replaces phrase-structure rules of all kinds, including the X′-theoretic ones, as the generative component of the theory. We will see that c-command can be directly derived from the effects of Merge. We will also see that Merge can give us a notion of projection. We look at the relation between Merge and LCA, and also introduce the Labelling Algorithm.

In this chapter, we introduce the concept of phases, a further development of the islands/subjacency/barriers line of investigation, but with many other consequences. We look at the notion of phase and the Phase Impenetrability Condition, in particular Chomsky’s original rigid definition of phases as CP, v*P and DP, which contrasts with Bošković’s contextual definition. We also look

at the consequence of the PIC that successive-cyclic movement has to pass through SpecvP and adduce a range of cross-linguistic evidence in support of this. We then turn to the question of the driver for successive-cyclic movement. The Labelling Algorithm (LA) can provide an elegant account of this. Finally, we see the evidence for a new set of islands and how the contextual definition of phases, the antilocality condition on movement and the PIC conspire to give a narrow window of movement-targets.

Here the focus is on covert cases of wh-movement, i.e. cases where the movement takes place in such a way that it cannot be directly observed in the output of PF, but only in terms of its effects on the semantic interpretation. The best-known example of this kind of wh-movement is found in Mandarin Chinese; accordingly we focus on that language. Next, we look at cases of covert movement in English: Quantifier Raising and wh-movement in multiple questions. Then we turn to the nature of the copies of movement, showing how copies can provide an account of reconstruction of binding-theory relations at the CI interface, as well as of partial movement and doubling at PF.

In this chapter we continue our investigation of hierarchical structure by focusing on the structure of TP and VP, concentrating on the latter. We first look at the structure of the clause. We then turn to the evidence that the subject is generated inside the VP (the VP-internal subject hypothesis, VISH) and raises to SpecT′ in English and many other languages. This leads to further discussion and examples of raising. Finally, we further elaborate the structure of VP, introducing VP-shells, structures where one VP is embedded in another.

In this chapter we continue our investigation of hierarchy by looking at head-movement, i.e. how heads of phrases may move and combine. In addition to seeing how this kind of movement works in technical terms, and what the empirical motivation for it is, we also introduce a major locality condition, the Head Movement Constraint. Then we return briefly to the topic of passives, and introduce raising. Finally, we make a first attempt at formulating a general notion of locality which unifies the Head Movement Constraint with a locality condition applying to both passives and raising.

In this chapter we observe that syntax is mostly silent; given the overall organisation of the grammar, there are good reasons to expect this to be the case. Furthermore, among the silent elements there are, in addition to copies, empty pronouns and covert movement, various kinds of ellipsis. VP or predicate-ellipsis is quite rich in English, while NP-ellipsis is meagre. Ellipsis displays a number of departures from absolute identity of the antecedent and elided constituent, notably but not only sloppy readings and voice mismatches. We also look at the distinction between deep and surface anaphora and, following on from this, evidence that radical prodrop in East Asian languages appears to involve NP- or argument-ellipsis.

This chapter looks at the syntactic, i.e. phrase-structural, definitions of grammatical functions put forward in Chomsky (1965), which we restate using X-bar theory. We then submit these definitions to the ‘relational-grammar critique’, to adopt a term coined by Baker (2001), which suggests not just that Chomsky’s definitions are incorrect, but that something closer to the traditional idea that grammatical functions are primitives of syntactic theory is the right approach. One aspect of this critique is that constituency tests do not give clear results in many languages (English being something of an exception). Instead, we propose that asymmetries in c-command relations can provide us with a more reliable and general guide to constituency, and hence phrase-structural relations. This allows us to maintain a configurational definition of grammatical relations. In the final section of this chapter, we look at a construction which appears to centrally involve grammatical functions: the passive. We will see how the passive can be elegantly and usefully defined in purely phrase-structural terms. The conclusion is that grammatical relations can be reduced to phrase-structural relations, and as such are not primitive elements of syntactic theory. This is an important step in establishing the primacy of configurational, hierarchical, phrase-structural relations.