Functional shift, also known as conversion or zero-derivation, is a word-formation process that converts a word from one syntactic category (part of speech) to another with no change to the form of the word (e.g., Brinton & Brinton, 2010; Quirk et al., 1985). For example, the word bride is commonly a noun as in the following excerpt (1a) from Shakespeare, and it is converted to a verb with no visible morphological changes in (1b).

“Neighbours and friends, though bride and bridegroom wants/For to supply the places at the table,/You know there wants no junkets at the feast.”

“Shall sweet Bianca practise how to bride it?”

(Shakespeare, The Taming of the Shrew, Act III, Scene ii)

As shown in (1), whether the word bride functions as a noun or a verb, it refers to two related senses, which makes the two identical forms with different functions polysemous. Typically, the syntactic context offers reliable cues for disambiguation between the grammatical functions of a polysemic word.

Beyond Shakespeare’s practice, functional shifts are a surprisingly common type of word formation in modern English (Bauer, 2004), as evidenced by our frequent encounters with words such as hammer, sneeze, and dance, among others (Farrell, 2001).

Although there have been abundant theoretical studies on functional shift to answer the question of whether it is a derivational or inflectional process (see Bauer & Valera, 2005 for a review), very little psycholinguistic research has been conducted on how it is processed in real time. Moreover, most studies investigated the processing of functional shift in the first language (L1), while research comparing its psycholinguistic aspects in the L1 and the second language (L2) is extremely rare (e.g., Bordag & Opitz, 2022; Opitz & Bordag, 2020; Stolterfoht et al., 2010). In particular, to our knowledge, there are no psycholinguistic studies on the processing of functional shift in L2 learners of English with a non-Indo-European L1 background (e.g., Korean or Chinese).

In the following sections, we first review the findings concerning the processing of functional shift. Next, we present our research questions and predictions regarding L2 processing of functional shift based on the findings reported in the literature.

Despite the high frequency of functional shift in English, psycholinguistic research concerning its processing is sparse (e.g., Lukic et al., 2019; Pliatsikas et al., 2014). However, recent neuroimaging studies have shown that polysemic words (e.g., the/to brush), which have more than one grammatical function, are processed differently from unambiguous (e.g., the song, to send) or homonymic words (e.g., to bear). Some functional magnetic resonance imaging (fMRI) studies have demonstrated that polysemic words increased neural activity, as compared to unambiguous words, within the left posterior inferior frontal gyrus (Burton et al., 2009; Gennari et al., 2007; Pliatsikas et al., 2014), reflecting the complexity of the conversion process between the two functions of a word. Using a grammaticality judgment task, Lukic et al. (2019) also reported longer response times and greater activation in the left inferior frontal and parietal gyri for polysemic verbs. These studies argued that the increased neural activation reflects the cost of the grammatical word formation that converted the noun (e.g., the brush) into a verb following the infinitive to.

Given that the polysemic words tested in these fMRI studies are balanced in terms of their frequency for each grammatical function, it is interesting that the converted words still triggered a grammatical process (i.e., conversion) in online processing. These findings raise the question of how “Shakespearean” functional shift (i.e., word class conversion as used by Shakespeare, as in the case of “companion” in you would companion me) would be processed online. Shakespearean functional shift refers to an unexpected use of a word in a sentence in terms of its grammatical status, which Thierry et al. (2008) claim induces a surprisal effect on the brain and thereby makes reading Shakespeare dramatic.1 This question has rarely been explored to date; to our knowledge, Thierry et al. (2008) and Keidel et al. (2013) are the only two studies that investigated the processing of functional shift in an experimental setting with neural evidence. Thierry et al. (2008) directly addressed the issue of whether the functional shift is processed as a syntactic violation in real time. In an event-related potential (ERP) experiment, they tested the effect of Shakespearean functional shift on meaningfulness decisions to sentences containing a functional shift, a semantic incongruity, a double violation, or no violation. The results revealed that the Shakespearean functional shift engendered significant left anterior negativity (LAN) and P600 modulations. These ERP components indicate that the native speakers of English took the grammatically shifted word as a syntactic violation upon encountering it in online processing. Interestingly, Shakespeare himself selectively used functional shift as a tool to work against the laws of grammar (Blake, 1983), which leads to a type of “neurological tempest” (Keidel et al., 2013).

In an investigation of native speakers’ semantic judgments using fMRI, Keidel et al. (2013) supported Thierry et al.’s (2008) conclusion that Shakespearean functional shift involves the detection of syntactic violations. Keidel et al. (2013) demonstrated increased activation for sentences containing a grammatically shifted word in several brain regions typically associated with grammatical processing (i.e., bilateral IFG) in general, the generation of the P600 (i.e., the basal ganglia, in line with Frisch et al., 2003), and error detection (i.e., the anterior cingulate cortex and the superior frontal gyrus, corroborating the findings of Bush et al., 2000; Dosenbach et al., 2007).

Currently, there is hardly any psycholinguistic research concerning the processing of functional shift in L2, apart from Bordag and Opitz (2022) and Vilas et al. (2019). Bordag and Opitz reported on a direct comparison between native German speakers (L1) and Czech learners of German (L2) focusing on the processing of conversion in German. Using visual priming experiments, they found that a converted word (e.g., das MIETEN “the renting”) had a partial priming effect on the processing of a form-identical word (e.g., wir MIETEN “we rent”). Based on this priming effect, they argued that form-identical words with different grammatical functions share the same entry in the mental lexicon. Crucially, there were no significant differences in the priming pattern between their participants’ L1 and L2 in terms of both accuracy and response times. The authors interpreted these results as evidence that the processing of conversion is not fundamentally different in L1 and L2 (at least at the advanced or upper-intermediate level of L2 proficiency).

Vilas et al. (2019) focused on the effect of age of acquisition (AoA) on the processing of functional shift. Through observing ERP amplitudes and time-frequency modulations, they found a main effect of AoA on the processing of the Shakespearean functional shift as reduced fronto-posterior theta-band oscillations across semantic (200−500 ms) and morphosyntactic integration (500−800 ms) time windows. That is, only early bilinguals, not late bilinguals, were sensitive to the anomaly of shifted words as observed in native speakers. Thus, the results clearly suggest that L2 learners, especially those who had begun acquiring the L2 at a relatively later age, would process the Shakespearean functional shift in a different way from natives. Even highly advanced L2 learners may not show the same brain potential as the Shakespearean functional shift (i.e., P600) that was observed in the L1 studies (Keidel et al., 2013; Thierry et al., 2008).

Unlike Bordag and Opitz’s (2022) results, Vilas et al.’s (2019) findings are more closely in accord with many other L2 studies that generally report differences between the processing of ambiguous words in L1 and L2. For example, while homonymous words are associated with greater processing cost even in L1, ambiguity processing is less accurate and engenders greater processing time in L2 than in L1 (e.g., Degani et al., 2014; Eddington & Tokowicz, 2015; Lukic et al., 2019). In relation to this study, of particular interest is the L1–L2 difference in the processing of unbalanced ambiguous words whose two meanings are not balanced in frequency or dominance (e.g., a bank: financial institution as the primary meaning, the edge of a river as the secondary meaning). Most studies of L1 processing have shown that all meanings of an unbalanced ambiguous word are co-activated initially, and the context-inappropriate secondary meaning is suppressed at a later stage (e.g., Duffy et al., 1988; Ihara et al., 2007; Love et al., 2003; Onifer & Swinney, 1981; Seidenberg et al., 1982; Swaab et al., 2003; Swinney, 1979). L2 processing of ambiguous words differs in various ways from L1 processing. Some L2 studies demonstrated simultaneous activation of both meanings but no suppression of the context-inappropriate secondary meaning (e.g., Elston-Güttler & Friederici, 2005; Love et al., 2003). Other studies show that only the primary meaning is activated initially and the contextually appropriate secondary meaning (e.g., a bank as an edge of a river) is activated subsequently (e.g., Quarbi, 2022).

Currently, little is known about whether the Shakespearean functional shift is recognized and processed as a syntactic violation or a semantic anomaly in L2. In this context, this study aims to investigate how Korean learners of English recognize a word when its functional status is shifted and thus unexpected. More specifically, using an ERP paradigm, we asked whether upon encountering the Shakespearean functional shift, Korean learners of English would process it as a syntactic violation as the native speakers did in Thierry et al. (2008) or as a semantic anomaly.

If Korean learners of English process functional shifts as a syntactic violation as natives do, LAN and P600 effects will appear (e.g., Barber & Carreiras, 2005; Coulson et al., 1998; Friederici, 1995, 2002; Osterhout & Holcomb, 1992). The LAN effect is claimed to reflect the detection of a morphosyntactic violation (Barber & Carreiras, 2005; Coulson et al., 1998; Friederici, 1995, 2002). The second syntax-based component, P600, a positive-going waveform peaking around 600 ms after the onset of a word, is associated with the detection of syntactic violations (e.g., Osterhout & Holcomb, 1992) as well as with later syntactic reevaluation process or structure reanalysis and is sensitive to the failure of syntactic integration (Hagoort et al., 1999; Kaan et al., 2000; Münte et al., 1998). In contrast, if Korean learners of English process the functional shift as a semantic anomaly, the N400 effect will appear. The N400 is a negative-going deflection with an average peaking latency of 400 ms post word onset, typically elicited by semantic violations (Kutas & Hillyard, 1980, 1984; Osterhout & Nicol, 1999) and/or difficulty in semantic integration (Berkum et al., 1999; Hagoort et al., 2003).

Comparing this study’s results with Thierry et al.’s (2008), we may add empirical evidence regarding whether L2 processing is fundamentally different from L1 processing, particularly in terms of semantic and/or syntactic representations. If L2 processing of functional shift elicits the same or similar ERP components as compared to L1 processing, this indicates that there is no fundamental difference between L1 and L2 (e.g., McDonald, 2006). Conversely, if L2 processing of functional shift is distinct from L1 processing, as reflected in their ERP modulations, this indicates that L2 processing is different from L1 processing, as claimed in many previous L2 studies (e.g., Clahsen et al., 2010; Heyer & Clahsen, 2015; Jacob et al., 2018; Kırkıcı & Clahsen 2013; Neubauer & Clahsen, 2009; Silva & Clahsen, 2008; Veríssimo et al., 2018). This difference between L1 and L2 could be explained by the Shallow Structure Hypothesis (SSH), which postulates that L2 speakers’ manipulation of abstract syntactic representations in real time is less robust than L1 speakers’ (Clahsen & Felser, 2006).