The multiple-lemma representation of Italian compound nouns: a single case study of deep dyslexia

It is not clear how compound words are represented within the influential framework of the lemma-lexeme theory. Theoretically, compounds could be structured through a multiple lemma architecture, in which both compound and constituent lemma-nodes are involved in compound processing. If this were the case, syntactic properties of both the compound and its constituents should play a role when performing tasks involving compound processing, e.g., compound-word reading. This issue is investigated in the present study through an assessment of the performance of a deep dyslexic patient (GR) in three compound-reading experiments. In the first experiment, Verb-Noun (VN) compound nouns (e.g., lavapiatti, “dishwasher”, lit. wash-dishes) were employed as stimuli, while in the second, VN compound stimuli were embedded in sentences, and were compared to paired verb phrases (e.g., lui lava piatti, “he washes dishes”). Position-specific effects were ruled out by means of a third experiment, which investigated the retrieval of Noun-Noun compounds (e.g., pescespada, “swordfish”, lit. fishsword). In experiment 1, GR made errors on the verb constituent more frequently than on the noun, an effect that did not emerge in Experiment 2: when embedded in sentences, VN compounds were read significantly better than verb-phrases and no grammatical-class effect emerged. In Experiment 3, the first and the second constituent were read with the same level of accuracy. The disproportionate impairment, which emerged in reading the verb component of nominal VN compounds, indicates that the grammatical properties of constituents are being retrieved, and thus confirms access to the constituent lemma-nodes. However, the results suggested a whole-word representation when compounds are embedded in sentences; since the sentence context affects the access to compounds through syntactic constraints, whole-word representation is arguably at the lemma level as well (multiplelemma representation). Experiment 3 indicates that these effects cannot be accounted for by a position-specific impairment.