The Constraints of Dual Morphological Systems on Visual Word Processing in Maltese

Abstract

In this dissertation, I conduct a visual lexical decision megastudy for the Semitic language Maltese and describe the construction of “MaltLex”, a database of visual lexical decision responses to 11,000 real words and 11,000 non-words by native Maltese speakers. MaltLex is the first such megastudy for a Semitic language, which characteristically use nonconcatenative morphological processes: Nonconcatenative morphology poses novel challenges for theories of word recognition and morphological decomposition which may be explored through the MaltLex dataset. Maltese is also unique among Semitic languages in several respects, such as in its use of the Latin alphabet and in its split lexicon (roughly half of all Maltese words are of non-Semitic origin and use only concatenative morphological processes), and so Maltese provides unique testing conditions for the role of letters, morphology, etc. in visual word processing. I conduct several “virtual experiments” (i.e. where I analyze subsets of the MaltLex dataset) in order to (1) validate MaltLex by replicating well-established findings from the lexical decision literature and (2) build upon our understanding of lexical processing in Maltese and Semitic languages by demonstrating several novel findings. In Chapter 1, I explore how word frequency and individual differences in language use and proficiency mediate lexical processing in Maltese. I replicate the canonical word frequency effect (readers judge more frequent words faster and more accurately) and show that a frequency measure that is based on the number of documents in which a word occurs in a corpus (contextual diversity) better predicts lexical processing performance than does a more traditional measure that is based on a word’s total number of occurrences (word frequency), which is consistent with megastudy-based research on non-Semitic languages (e.g. Brysbaert and New 2009). All MaltLex participants were bilingual in Maltese and English, and I also show that the more English-dominant participants exhibited smaller processing differences in judging the lexicality of non-Semitic versus Semitic Maltese words than did the more Maltese-dominant participants, which is broadly consistent with “cognate advantage” effects found in multilingual lexical processing (e.g. Poort and Rodd 2017). In Chapter 2, I explore how the orthographic form of a word mediates lexical processing in Maltese. I replicate standard orthographic neighborhood density effects: MaltLex participants judge real words that have more orthographic neighbors faster and more accurately than those with fewer neighbors, though this advantage becomes smaller as words increase in frequency; and they judge non-words with more neighbors slower and less accurately than those with fewer neighbors (e.g. Andrews 1989). Beyond the standard neighborhood density effect, I also show that MaltLex participants are even slower and less accurate at judging the lexicality of non-words that differ from an existing Maltese word except by the addition or omission of a diacritic, suggesting that Maltese readers process certain pairs of letters alike during lexical processing (e.g. “g” and “ġ”). In Chapter 3, I explore the role that the consonant letters that comprise individual words play in visual lexical processing in Maltese. In a novel visual masked priming study, I show that similar priming effects obtain for non-Semitic versus Semitic words when primed by strings that consist of their consonant letters (e.g. pnġ priming PINĠA ‘to paint’), which is consistent with subset priming effects in non-Semitic languages (e.g. Anderson and Geary 2018, Duñabeitia and Carreiras 2011) but inconsistent with Geary and Ussishkin (2018), who found such priming effects for Semitic Maltese words but not non-Semitic Maltese words. I then analyze the MaltLex dataset, showing that the frequency of a word’s set of consonant letters across the Maltese lexicon mediates unprimed lexical processing and thus may mediate these differences in observed priming effects. I have hardly scratched the surface of the range of analyses that may be performed using the MaltLex dataset, which I make freely available to other researchers. I conclude by describing the structure of the MaltLex dataset and how other researchers may access it in Chapter 4.