Beginners Read Aloud : High versus Low Linguistic Levels in Swedish Beginners' Oral Reading

Sammanfattning: The aims of this thesis were to examine the utilisation of various linguistic levels in the oral reading of running texts among Swedish beginning readers, and specifically to question the supposedly predominant role of lower (i.e. sub-lexical) linguistic levels by also examining possible evidence of the utilisation of information at the syntactic or semantic levels, as well as textual context. The investigation is based on a corpus constructed from the oral reading of running texts and includes a number of studies using both quantitative and qualitative error analyses.The analyses confirm that other linguistic levels than the sub-lexical have an impact on reading. This was shown both in the linguistic acceptability of errors and the extent to which errors were corrected depending on linguistic acceptability. Although the natural point of departure seemed to be the graphemic level, analyses revealed that graphemic complexity or word transparency alone could not explain error frequencies. In quite a few cases, qualitative analyses revealed, for instance, that higher linguistic levels or knowledge of the world could explain both why words did and did not result in reading errors. However, phonological quantity appeared to be a major difficulty throughout the study, which is clearly related to the graphemic or phonological level.Some differences regarding the developmental perspective were observed. One study indicated that the readers might develop stepwise regarding their utilisation of various linguistic levels, in the sense that they appeared to rely mainly on lower linguistic levels early in reading development. Later they seemed to be dependent on higher linguistic levels, and ultimately they seemed to be sensitive to, rather than dependent on, higher linguistic levels.An interesting result was that the readers seemed to use different strategies for different kinds of words throughout the investigation, using a direct decoding strategy for frequent words, but using a letter-by-letter decoding strategy for less frequent or graphemically complex words.