Modality and morphology: what we write may not be what we say

Abstract

Written language is an evolutionarily recent human invention; consequently, its neural substrates cannot be determined by the genetic code. How, then, does the brain incorporate skills of this type? One possibility is that written language is dependent on evolutionarily older skills, such as spoken language; another is that dedicated substrates develop with expertise. If written language does depend on spoken language, then acquired deficits of spoken and written language should necessarily co-occur. Alternatively, if at least some substrates are dedicated to written language, such deficits may doubly dissociate. We report on 5 individuals with aphasia, documenting a double dissociation in which the production of affixes (e.g., the -ing in jumping) is disrupted in writing but not speaking or vice versa. The findings reveal that written- and spoken-language systems are considerably independent from the standpoint of morpho-orthographic operations. Understanding this independence of the orthographic system in adults has implications for the education and rehabilitation of people with written-language deficits.

Keywords: cognitive neuroscience; language; psycholinguistics.

© The Author(s) 2015.

Figures

Figure 1

Accuracy with verb inflections (left) and noun inflections (right) in spoken and written production. While inflections were produced more accurately in speaking than writing by KSR, PW, DHY and AES, a complementary pattern was exhibited by VBR. See Table 2 for details of the statistical evaluation.

Figure 2

Distribution of errors across serial position for multi-morphemic and length-matched mono-morphemic errors. Segment accuracy is scored in three bins, representing (for multi-morphemic words) the first half of the stem (1st), the second half of the stem (2nd) and the inflection (3rd/inflection). Mono-morphemic error binning corresponded to the same serial positions as in the matched multi-morphemic items. The error bars reflect bootstrapped 95% confidence intervals around the mean for the mono morphemic error distributions. Written responses are presented for AES (a), DHY (b), KSR (c), and PW (d); spoken responses for VBR (e). For all participants, errors occurred significantly more frequently in final positions for multi-morphemic versus mono-morphemic words (p < .0001).

Figure 3

The number of morphological errors actually produced by DHY (black bar) and the distribution of the numbers of morphological errors expected by chance for DHY’s data set, as determined by 10,000 runs of a Monte Carlo analysis (white bars). The results reveal that DHY’s morphological errors were significantly more frequent than expected from a random sampling of orthographic neighbors (p < .0001). Similar distributions were obtained for the other participants, and findings are reported in Table 4.