Dissertations, Theses, and Capstone Projects

Date of Degree


Document Type


Degree Name



Speech-Language-Hearing Sciences


Valerie Shafer


Loraine Obler

Committee Members

Mira Goral

Niels Schiller

Subject Categories

First and Second Language Acquisition | Psycholinguistics and Neurolinguistics


Bilingualism, multilingualism, codeswitching, sentence-processing, EEG, neurolinguistics


Neurophysiological sentence processing studies are inconsistent about the additional costs that language switching within sentences (i.e., codeswitching) may bring about (cf. Valdés Kroff et al., 2020; Yacovone et al., 2021). There is discussion about whether there are, in fact, additional costs and, if so, about the origins of those costs, since some findings are consistent with effortful processing when comprehending single-language (not switched) words observed in monolingual studies. Specifically, studies are divided on whether codeswitched words are more difficult to lexically access than single-language equivalents of similar semantic predictability, as indexed by an increased N400 effect. The same studies, however, are showing that codeswitches typically elicit a large posterior positivity (referred to as the LPC), hypothesized to index sentence-level reanalysis. This response is typically not elicited in single-language conditions, unless in the case of anomalies, but it is unclear how this processing compares to that of codeswitches.

This dissertation examined the extent to which codeswitch costs are explained within a general (monolingual) sentence processing framework. Codeswitches were contrasted with lexical expectancy in a group of Dutch-English multilingual adults (n = 34), specifically comparing NonSwitched vs. CodeSwitched and Expected vs. Unexpected lexical completions using spoken sentences. In addition, a NonSwitched-Anomalous condition was included to contrast sentence-level reanalysis elicited by anomalies (single-language) versus codeswitches (non-anomalous). This condition employed sentences like Oma kan niet goed zien zonder haar ... (‘Grandma can’t see properly without her ...’) where bril (‘glasses’) is the NonSwitched-Expected completion, lenzen (‘contacts’) is the NonSwitched-Unexpected completion and its CodeSwitched equivalents were glasses and contacts. By contrast, taart (‘cake’) was an example of a NonSwitched-Anomalous completion.

The results indicate that, relative to the predicted single-language completions (NonSwitched-Expected), lesser predicted codeswitched completions (CodeSwitched-Unexpected) increased the N400 response. Translation equivalents of predicted single-language completions (CodeSwitched-Expected), however, did not increase the N400 response. Single-language anomalous completions (NonSwitched-Anomalous) elicited the greatest increase in N400 response, including relative to less well predicted codeswitched completions (CodeSwitched-Unexpected). In terms of sentence-level reanalysis, single-language anomalies (NonSwitched-Anomalous) and codeswitches, irrespective of their lexical expectancy, elicited a late positivity (LPC). The positivity increase did not differ between CodeSwitched and Anomalous completions, whereas both the CodeSwitched and the Anomalous completions differed from NonSwitched completions.

These results align with studies that observed no additional costs for lexically accessing codeswitches. Specifically, lexical access costs brought about by codeswitches can be explained in terms of lexical expectancy, where translation equivalents of predicted single-language completions are not more difficult to lexically access compared to their single-language counterparts of equal semantic probability. Translation equivalents of less well predicted single-language completions, however, are more challenging to lexically access. But none of the non-anomalous completions were as difficult to lexically access as the anomalous single-language completions. In addition, in line with the majority of the intrasentential codeswitching studies, this dissertation observed neural responses (LPC) that may reflect sentence-level reanalysis. These findings specifically add that sentence-level reanalysis for codeswitches did not differ depending on their lexical expectancy, despite having observed this difference during lexical access. This dissertation adds that sentence-level reanalysis did not differ in terms of degree between anomalous and codeswitched completions.

These findings, therefore, suggest that codeswitching costs can be partially explained within a general sentence processing framework, where lexical access of codeswitched words can be facilitated when the codeswitch aligns with the predicted lexical completion. However, a general sentence processing framework cannot account for codeswitches resulting in sentence-level reanalysis, since these responses were only observed t0 single-language anomalous continuations. Lastly, irrespective of the observation that codeswitches may result in sentence-level reanalysis, this dissertation findings are consonant with a model of language non-selective access, which allows for rapid lexical access of words from the non-target language within a target language sentential context and rapid sentence-level reanalysis of the input and the prior context.