Second language phonology: The phonetics–phonology divide,its underrepresentation in L2 research,and the richness of phonological representations from segments to prosodic structure

1 Underrepresentation of L2 phonology

Archibald’s keynote highlights the underrepresentation of L2 phonology within generative second language acquisition (SLA) literature. While this is already widely recognized, the author’s purpose is not simply to describe or reemphasize the limitations of current approaches, but rather to actively show that by overlooking the learnability of abstract mental representations in our investigations of L2/L3 environmental input and output, we fail to grasp critical insights into the true nature of multilingual grammar. The author argues compellingly that insights from L2 phonology are as valuable as those derived from morphosyntax in understanding the language faculty and the process of L2 acquisition. This calls for a re-evaluation of the focus within SLA research, advocating for a more integrated approach that considers phonological aspects as fundamental to language learning. As the author illustrates, phonology is not, after all, different from morphosyntax in its complexity and in illustrating abstract linguistic representations. The author does this not just by pointing to relevant research in prosody and its acquisition, but also by illustrating other domains, such as syllables and segments, though the focus still remains on the components of prosody largely, including the foot, the prosodic word and higher-level constituents of the Prosodic Hierarchy, such as the phonological phrase.

2 Distinction between phonetic and phonological studies

The keynote emphasizes the necessity of distinguishing between the wealth of phonetic studies present in the L2 literature and the comparatively limited phonological research in this field (despite the recent activity in theoretical phonology). The former primarily concerns the physical properties of speech, as the author maintains, while the latter addresses the abstract linguistic representations underlying sound systems, feet, words and phrases. This distinction is crucial for framing future research agendas and ensuring that phonological insights are adequately integrated into the broader conversation on L2 acquisition, a goal that the author achieves skillfully.

In addition, the author also argues that the two, phonetics and phonology, also differ in that the former is linear while the latter is recursive. It should be noted, however, that although phonology is certainly not simply linear in the sense of phonetics, as it has hierarchical structure that could be quite complex, the claim that it is recursive is not a widely accepted view, even in formal phonological literature. Nevertheless, it remains true, as the author argues, that phonetics is primarily concerned with physics and muscle movements, while phonology centers on cognition and the linguistic faculty, with phonology being inherently hierarchical. However, as one reviewer observes, phonetics involves continuous variables, whereas phonology typically deals with categorical variables, with the physics and muscle movements underpinning the continuous ones. This perspective suggests that framing the distinction as linear versus hierarchical may not fully capture the complexity of the relationship between these two domains. Connected to this point, it is also worth noting, as one reviewer suggests, that phonology has increasingly been modeled as gradient or continuous in recent years, influenced by probabilistic approaches (e.g. Albright and Hayes, 2003; Flemming, 2001; Hayes, 2000; Pierrehumbert, 2003), further complicating the assumption that phonology is strictly nonlinear or categorical. Having said all this, and returning to the issue of recursion and complexity, it is important to note that phonology can exhibit a complexity comparable to syntax without necessitating recursion; in fact, certain aspects of phonology may even surpass syntax in complexity. For example, phonology operates across multiple representational layers – such as in tonal spreading and vowel harmony (see, for example, Feature Geometry; Clements, 1985; Clements and Hume, 1995; Özçelik, 2024) – a characteristic absent in syntax. This multi-tiered structure and hierarchical nature play a crucial role in evaluating phonological complexity and suggest that recursion may not be a prerequisite for such sophistication.

Given that phonology is part of the grammar and that phonetics is more physical in nature, it is perhaps not surprising that native-like phonology is quite common among language learners, regardless of the subarea of phonology one examines, assuming access to Universal Grammar (UG) in that domain (Archibald, 1993, 2022, 2024; Özçelik, 2016, 2018; Özçelik and Sprouse, 2017), while native-like accent, affected by established physical habits and muscle movements, is less common. Regardless of the debate on recursion then, the author’s main point remains the same, i.e. that ‘phonological representations and processes cannot be reduced to phonetic properties,’ and ‘algebraic representation’ and ‘structure dependency’ are entailed for phonology, while learning based on analogy alone (which, for example, Exemplar-based models use) does not work. In other words, the point is, as the author argues, ‘syntax/phonology is a cognitive interface; it’s not just a motoric interface.’ The author conveys this message in utmost clarity, which will help in rectifying the situation for the field of L2 phonology described in Section II.1.

An additional consideration may be necessary here: both phonology and syntax extend beyond general cognitive principles, as certain production or acquisition patterns cannot be fully accounted for even by cognitive processes, for two reasons: First, continuous phonetic variables play a crucial role in binary phonological analysis. While stress, and prosody more broadly, is a core area of phonological investigation, its correlates – pitch, intensity, and duration – are inherently continuous. There is no fixed threshold for how high a pitch must be for a syllable to be considered stressed or how loud it must be to qualify as stressed. Instead, these features are relative, depending on the pitch and loudness of neighboring syllables. Consequently, stress determination lacks a clear-cut boundary between stressed and unstressed syllables; instead, it relies on continuous variables in the speech stream, although its end result is binary (i.e. a stressed vs. unstressed syllable), along with the parameters or constraints that lead to these surface variations (e.g. left- vs. right-headedness, iterativity vs. non-iterativity, etc.).

Second, and equally importantly, language learners exhibit patterns that cannot be accounted for by cognitive principles alone. In fact, under certain well-defined conditions, linguistic principles directly conflict with cognitive principles, further underscoring the limitations of a purely cognitive approach to language acquisition. This is evident in both L2 morphosyntax (e.g. White, 2003) and L2 phonology (e.g. Özçelik, 2018), where certain phenomena resist explanation solely through cognitive frameworks. Although this point is alluded to in the keynote, by attracting attention to domain-general vs. domain-specific approaches (see Section II.5), it risks being overshadowed by the author’s pronounced emphasis on phonology (alongside morphosyntax) as inherently being cognition. For instance, studies on L2 acquisition of word stress have shown that certain cognitively plausible acquisition stages – such as the production of weight-insensitive iambs by English-speaking learners of Turkish, a stage that would plausibly follow one in which these learners switch from (weight-sensitive) trochees to (weight-sensitive) iambs – do not occur because they are ruled out by principles of UG. This holds true despite input and cognitive factors that might otherwise favor such interlanguage grammars. These factors include the predominance of word-finally prominent words in Turkish, pedagogical approaches emphasizing word-final stress (even in the presence of a non-final long syllable, e.g. [vaːlí] ‘governor’), and the inherent difficulty of eliminating metrical feet from the learner’s grammar. Yet, as Özçelik (2018) demonstrates, weight-insensitive iambs (e.g. [(vaːlí)_Ft]) are notably absent in interlanguage grammars, highlighting the constraints UG imposes on acquisition beyond what cognitive and environmental influences would predict. Thus, while phonology is indeed more closely aligned with cognitive processes than with physical ones, as the author suggests, it nonetheless relies on physical and continuous correlates. At the same time, it cannot be reduced to cognition alone, as both its nature and acquisition require explanations that extend beyond cognitive principles.

3 Factors contributing to limited attention on L2 phonology

Archibald explores various reasons for the historical neglect of L2 phonology, most notably the influence of the popular Minimalist approaches to syntax (e.g. Chomsky, 1995), for which phonology is unjustifiably reduced to merely an interface to the motor system, and the prevalence of oversimplified concepts like ‘input noticing’ widespread in the non-linguistic approaches to second language learning. These factors have led to a narrow focus that often overlooks the complexities of phonological systems and their acquisition, such as the inadequacy of input alone to account for the complex knowledge that L2 learners acquire. Conversely, in some cases, learners encounter difficulties in specific domains even when robust input is available. By addressing issues like these, Archibald invites a more nuanced discussion that acknowledges the rich and multifaceted nature of phonological knowledge.

The author also blames Optimality Theory (OT) (Prince and Smolensky, 2008), stating that it has diverted attention from representational issues as it is a theory of computation. While OT’s computational focus is undeniable, it works in tandem with theories of representation, including those developed in the 1970s and 1980s, such as Autosegmental Phonology and Feature Geometry (see, for example, Clements, 1976, 1985; Clements and Hume, 1995; Goldsmith, 1976). None of these have received enough attention in the L2 acquisition literature, despite being representational (see Özçelik and Sprouse, 2017). In addition, OT’s focus on computation has resulted in numerous phonologically informed publications in first language acquisition, especially from a computational perspective, but not in L2 acquisition. Thus, contrary to the author’s suggestion, OT itself may not be at fault here.

Instead, it seems that Minimalist syntacticians and OT scholars have pursued separate directions, with Minimalist approaches ultimately gaining more prominence in generative studies of L2 acquisition. This preference is further supported by the long-standing – and, as the author effectively critiques, flawed – assumption equating phonology with phonetics, and in phonetics, pervasive non-native-like patterns are readily observed. This issue is compounded by the commonly held generative view that L2 acquisition involves access to Universal Grammar (UG), a concept that is challenging to substantiate in phonetics, given its basis in physics rather than within the language faculty.

4 Richness of phonological representations

As was also briefly mentioned in Section II.2, Archibald convincingly argues that phonology is as rich and complex a representational system as syntax, perhaps unlike phonetics. After all, both phonology and syntax share characteristics such as structure-dependence, hierarchical organization, and algebraic properties, suggesting that insights gained from phonology can inform learnability issues similar to those present in syntax. This parallel strengthens the case for phonological considerations in SLA research, highlighting the interconnectedness of language structures. Archibald further suggests that phonology is recursive, though this is an approach not widely accepted in the field of phonology, as mentioned in Section II.2, and many cases of what appears to be recursive structure can actually be accounted for by referring to iterativity, which also renders phonological grammars hierarchical. In other words, in order to reach the same conclusions as Archibald, one does not need to employ recursivity.

However, Archibald’s position that phonological systems are complex continues to be accurate, regardless. The trend in place in the L2 acquisition literature has not been reflective of this fact though. The author provides an excellent example from stress literature to illustrate this point: most of this literature treats stress as a property that either exists on a syllable or is absent (e.g. Dupoux et al., 1997), with no understanding of how various parameters or constraints work together to cause stress to fall on a specific syllable. This enduring issue exists despite the extensive body of formal phonological research on the topic since the 1990s (see, for example, Dresher and Kaye, 1990; Hayes, 1995). This approach, which remains quite prevalent in research on L2 acquisition of stress, not only overlooks the mechanisms by which stress is assigned to specific syllables, but also fails to adequately differentiate between concepts like stress and intonation. Such distinctions are crucial for accurately characterizing languages whose speakers have been loosely described as ‘stress-deaf’ in the acquisition literature, like French and Turkish, compared to those that are not, such as English and Spanish. This is because the underlying property that distinguishes the two types of languages is whether their native language is stress-based or intonational (see, for example, Özçelik, 2017, 2024, and the various chapters in van der Hulst, 2014).

Nevertheless, there have been some recent work, in particular starting with some of the author’s own (e.g. Archibald, 1992, 1993), which has been based on a deeper understanding of stress systems (see also Archibald, 1997, 1998; Garcia, 2020; Garcia and Goad, 2024; Garcia and Guzzo, 2017; Pater, 1997; Steele, 2002; Özçelik, 2016, 2018, 2021). The same trend is evident, to some extent, in other areas of phonology as well, other than word stress, starting from early 90s (e.g. Broselow, 1988; Broselow and Finer, 1991), although it has, thus far, been mostly limited to higher-level prosody, which is already quite like syntax in many ways and has significant interface with it (see, for example, work on the Prosodic Transfer Hypothesis by Goad and White, 2004, 2006, 2009, 2019).