2.1. Setting the stage

In the first chapter, ‘Overture’, Archibald sets out to argue for a unified, ‘internally consistent’ cognitive model of L2 phonology (p. 80). While he ‘builds on insights’ from Laboratory Phonology (Pierrehumbert et al. Reference Pierrehumbert, Beckman and Ladd2000) and Exemplar Theory (Johnson Reference Johnson, Solé, Beddor and Ohala2007) in developing his ideas (p. 2), his model departs markedly from these positions: in contrast to Laboratory Phonology and Exemplar Theory, phonetic details are not part of phonological representations in Archibald’s model (see §2.2). Phonetic details are relevant in his view, however, for explaining phonological learning (cf. Hale & Reiss Reference Hale and Reiss2000), and can account for gradual learning paths and gradient effects in acquisition or production.

A central theme of the monograph is the idea that L2 phonological learning presents two core difficulties that need to be explained (cf. Chomsky Reference Chomsky1986): Plato’s Problem, which is that L2 learners acquire linguistic knowledge that is impoverished or absent in the input (e.g., no direct acoustic cues), and Orwell’s Problem, which is that L2 learners fail to acquire linguistic knowledge despite its frequent occurrence in the input. Plato’s Problem can be exemplified by the phonemic contrast /p/–/b/, which learners need to determine is represented by the feature [voice] in French, but by the feature [spread glottis] in English (Iverson & Salmons Reference Iverson and Salmons1995). The monograph addresses this problem using abstract phonology and Distributed Morphology (DM; Halle & Marantz Reference Halle, Marantz, Hale and Keyser1993; Embick Reference Embick2010). Orwell’s Problem can be illustrated by the difficulty L1 French speakers have in acquiring the frequently occurring sound /h/ in L2 English (Mah et al. Reference Mah, Goad and Steinhauer2016). This issue is addressed through the concept of intake frequency (see also the discussion of Chapter 3 below in §2.2). Finally, Archibald delineates the three potential sources that are available to L2 learners for forming new L2 representations, all of which will play a role in the various chapters of the monograph: UG, input effects and L1 knowledge (i.e., transfer).

2.2. The grammar of phonology and the phonetics–phonology interface

In the second chapter, ‘Phonological Grammar’, Archibald lays out what he sees as the core properties and the design of IL phonological grammars. Crucially, Archibald argues that L2 phonology is part of cognition, that is, something that can be learned, rather than ‘merely motoric routines’ or ‘noticing’ phonology in the input (p. 79). What he means is that there is evidence that L2 phonological representations exhibit hierarchical representations, including features, syllables and feet, and that there is empirical evidence for these phonological categories in L2 grammars. Yet, these phonological features in the representations cannot be captured directly from the acoustic input.

Furthermore, Archibald argues against what he calls the ‘deficit hypothesis’ (p. 17), that is, the view that adult L2 learners are unable to acquire certain linguistic features in their L2 if they are not available in their L1. He proposes that L2 speakers can acquire novel L2 segments under certain circumstances, for example, very salient sounds with robust cues, such as a trilled [r] in Russian. (The notion of cue robustness is discussed further in Chapter 3.) He further sets out to clear up the confusion surrounding the concept of ‘stress deafness’. In an in-depth discussion, he argues that the term itself is inappropriate and misleading, and that it is crucial to distinguish between representations and processes. He asserts that stress is a complex system of knowledge and metrical representation that L2 speakers believed to be ‘stress-deaf’ can, in fact, acquire. Overall, he contends that IL grammars are constrained by the hypothesis space provided by UG.

These ideas about the IL phonological grammar are expanded upon in the third chapter, ‘The Phonetics/Phonology Interface’. In this core chapter, Archibald argues strongly for a separation between phonology – an abstract, categorical system – and phonetics, which is concerned with gradient and concrete aspects of speech. Importantly, for him, and following Ladd (Reference Ladd2014), gradient phonetic aspects should be part of processing and learning models rather than representational accounts. Drawing on evidence from L2 Japanese moras, illusory vowels and appendices, as well as feature geometries, Archibald puts forth the argument that IL phonological grammars – just like the grammars of L1 speakers – are best accounted for with symbolic systems that are not readily available in the auditory input, and hierarchical representations. He uses data to dispute Exemplar Theory (Pierrehumbert Reference Pierrehumbert, Bybee and Hopper2001) and ‘direct realist’ models of the phonetics–phonology interface (e.g., Brannen Reference Brannen2011; Wilson & Davidson Reference Wilson and Davidson2013), the latter of which would argue that listeners are capable of directly perceiving and thus representing phonetic details. Instead, he advocates an ‘indirect realist approach’, in which phonological features for representing sounds cannot be directly detected in the input, but must be learned.

While phonetic details have no place in phonological representations in Archibald’s model, he concedes that novel L2 structures can be acquired by either phonetic or phonological means. Archibald proposes that the ‘L1 filter’, which initially influences IL phonological grammars, can be ‘overridden’ under two types of conditions to allow new L2 representations to be established: in phonetics, where there is cue robustness in the input (cf. Wright Reference Wright, Hayes, Kirchner and Steriade2004; Keyser & Stevens Reference Keyser and Stevens2006), and in phonology, when L1 grammatical structures are redeployed for L2 representations. First, cue robustness – a type of saliency – is said to exist for sounds that are phonetically robust and for sounds in positions that enhance cues. Sounds with such properties are privileged in the phonological processing because they can pass through the L1 filter to become intake, thereby increasing their intake frequency. Here, Archibald discusses work on the perception of ejectives by L1 Spanish–L2 Yucatec Mayan learners (González Poot Reference González Poot2011). This study found that different types of L2 ejectives were more accurately discriminated in onset versus coda position: on the one hand, non-strident stops are privileged over strident affricates in onset position because of the more perceptible transitional release cues in the release bursts of non-strident stops; on the other hand, strident ejectives are privileged in coda position, arguably because strident ejectives have more robust cues than non-strident ones in coda positions. Intake frequency determines when (or whether) an L2 feature is phonologised, thereby accounting for the developmental sequence of L2 sound acquisition. Thus, intake frequency is proposed as an answer to Orwell’s Problem. A further argument is that segments that become intake faster are represented earlier in the underlying representation of lexical entries (p. 103).

The second option, redeployment of L1 knowledge, is a phonological way of acquiring novel L2 structures. Here, Archibald presents evidence from L1 English–L2 Japanese learners who were able to discriminate vocalic and consonantal length contrasts even though English lacks a length contrast for consonants (i.e., it has no contrastive geminates; Summerell Reference Summerell2007). Arguably, the L1 English speakers redeployed an L1 weight-bearing mora structure, that is, closed CVC syllables which are considered heavy because the final C projects a (weak) mora. With this and other examples, particularly on /sC/-clusters and illusory vowels, Archibald proposes that certain aspects of L1 knowledge can be transferred and used in novel ways for the L2, thus supporting the idea of full transfer in IL grammars.

2.3. The interfaces of phonological with morphological and syntactic grammar

In the fourth chapter, ‘The Phonology/Morphology Interface’, Archibald begins to argue for the crucial role of phonology within the larger context of morphological structures. He underscores that the interface between phonology and morphology has been relatively under-investigated in comparison to the interface between phonology and syntax. Thus, Archibald provides evidence from four phenomena to support the model he developed in previous chapters.

One morphophonological phenomenon discussed is the acquisition of an aspect of German plural allomorphs by beginner-level classroom learners (Archibald Reference Archibald, Leal, Shimanskaya and Isabelli2022). Specifically, he examines whether L2 German learners acquire the grammatical knowledge for German that the plural suffix -n cannot co-occur with umlaut, a phenomenon not typically explicitly discussed in German-as-a-foreign-language classrooms. Building on the work of Trommer (Reference Trommer2015, Reference Trommer2021), Archibald maintains that German has a single plural affix, an underspecified coronal node. According to this view, the co-occurrence of the suffix -n with umlaut would entail that the plural-marking coronal feature would have to be linked twice. This, however, would lead to a crossing of association lines, violating a core property of autosegmental phonology (Goldsmith Reference Goldsmith1976; Hammond Reference Hammond1988). Therefore, it is argued that the coronal feature can be realised as either a coronal consonant or a coronal vowel in German plurals. In a written plural elicitation task, Archibald (Reference Archibald, Leal, Shimanskaya and Isabelli2022) found that L2 German users made a large number of plural mistakes, yet rarely produced plurals with both the -n suffix and umlaut. Archibald thus argues that this indicates that the learners’ IL grammars adhere to the universal representational principle that disallows crossing association lines.

Another phenomenon discussed in this chapter is intra-word code-switching (ICS) in bi-/multilinguals, where the following patterns can be identified (e.g., Stefanich Reference Stefanich2019): while sentences can contain roots from either language, affixes are always drawn from the language of the syntactic structure, and ICS shows a single phonology, namely the phonology of the affix language. Archibald emphasizes the ‘explanatory adequacy’ of DM, as it can nicely account for the fact that roots in code-switches can be drawn from any language. In a bi-/multilingual context, this suggests that roots compete for insertion across languages, just as affixal allomorphs compete for insertion within a single language in DM. He further argues that the absence of phonological switching in ICS is consistent with the idea of derivation by phase (phonological persistence) as well as with Match Theory (Selkirk Reference Selkirk, Goldsmith, Riggle and Yu2011; Elfner Reference Elfner2015); the latter proposes an isomorphism between syntactic and prosodic structures, such that there is a (preferred) mapping between syntactic heads (X $^0$ ) and prosodic words (ω).

In the fifth chapter, ‘The Phonology/Syntax Interface’, Archibald extends his discussion of the central role of phonology for IL grammars to syntax. Here, he presents two interface examples in which he draws upon abstract hierarchical phonological – here prosodic – representations, for both the comprehension and production of syntactic structures. Archibald first discusses silent reading studies with people who are Deaf or Hard-of-Hearing (DH) and are, in part, L2 English users (Sagae Reference Sagae2007). Silent-reading studies capitalise on the idea that silent reading and inner speech involve phonological – including prosodic – activation, processing and encoding (e.g., Haber & Haber Reference Haber and Haber1982; McGuigan & Dollins Reference McGuigan and Dollins1989; Bader Reference Bader, Fodor and Ferreira1998).

Studies involving written garden-path sentences suggest that both participants with normal hearing and participants who are DH showed more difficulty (in an offline study) and more regression (i.e., fixation time in an eye-tracking study) for sentences requiring reanalysis; crucially, reanalysis is argued to be both syntactic and prosodic (cf. Bader Reference Bader, Fodor and Ferreira1998; Fodor Reference Fodor2002). It is concluded that both participants with normal hearing and people who are DH use prosodic encoding of sentences during silent reading, which explains their parsing and reanalysis patterns in the case of garden-path sentences. Archibald also takes this to mean that (i) cross-model redeployment of phonological representations is possible from a signed language to a spoken language, and that (ii) phonology must be represented cognitively, because the phonology of a signed language is not ‘articulatorily driven’ (p. 164). Note, however, that in the same research study, people who are DH did not perform like subjects with normal hearing on relative-clause attachment tasks, another prosody–syntax interface phenomenon (Sagae Reference Sagae2007). Here, it is suggested that the parsing procedures for L2 English relative-clause attachments could not be redeployed from the users’ L1 American Sign Language (ASL), and had not yet been learned by the L2 participants.

Second, Archibald discusses data on the production of WH-questions by L2 Japanese users. Here, the data are couched within Richards’s (Reference Richards2010, Reference Richards2016) Contiguity Theory, which states that contiguity between a WH-word and the interrogative feature [+Q] in WH-questions can be achieved either by surface linear adjacency in languages which move their WH-word to the specifier of CP, or else by prosodic contiguity in WH-in-situ languages. Prosodic contiguity for Japanese is defined as the absence of prosodic boundaries between the WH-word and a question particle [+Q], with a phonetic boost on the WH-word. To represent syntactic and prosodic tree structures of WH-questions in Japanese, Match Theory is again employed, which assumes an isomorphism between syntactic and prosodic structures and allows for recursive intervening phonological phrases after the WH-word (see Elfner Reference Elfner2015).

In a reported study (Archibald & Croteau Reference Archibald and Croteau2021), L2 Japanese users were shown to have acquired target-like phonology, specifically prosodic contiguity, in that they used relatively level pitch for the WH-word and the [+Q] particle, as well as an absence of pitch peaks for intervening items; at the same time, however, the L2 users demonstrated neither target-like phonetic realisation of the pitch boost on WH-words nor native-like F0 decrease on the words between the WH-word and the question particle. These data are presented as further evidence for the role of phonology at the interface with syntax, given that the L2 Japanese users displayed target-like phonology without phonetic ‘imitation’. Archibald takes this to indicate that IL grammars are constrained by properties of UG such as those proposed in Match Theory and Contiguity Theory, which together achieve a prosodic structure, here prosodic contiguity, that licenses WH-phrases in situ (p. 200). More generally, Archibald proposes this as evidence for hierarchical phonological and syntactic representations and against the Shallow Structure Hypothesis (Clahsen & Felser Reference Clahsen and Felser2006) in L2 grammars.

2.4. Comparisons with other new models

The monograph concludes with a short sixth chapter, ‘Underture’, in which Archibald recapitulates his indirect realist, integrated model of SLA. In doing so, he briefly touches upon other theories of SLA that make different assumptions. Without going into too much detail, he outlines his concerns with, for example, a Bayesian account of language acquisition (Wilson & Davidson Reference Wilson and Davidson2013). In essence, he argues that such approaches are not designed to model ‘the acquisition of knowledge’ but merely ‘courses of action or performance’ (p. 202; emphasis in the original).