Reserch in Spelling Reform

PART III

CHAPTER 10. CONCLUSION

1.  Summary of experimental findings

Experimental focus has been placed on Experiment 9.1, and some other studies in the network of experiments are reported in lesser detail.  Some studies remain at the pilot level, with trends in findings that await statistical support. Further  experiments and studies not essential to the body of the text are summarised in the Appendices, and some others are noted briefly so that their experimental designs may  be taken up and developed further. 

The advantages of the alphabetic principle for readers and learners

Demonstration experiment 6.1 shows the value of the alphabetic principle in reading unfamiliar vocabulary, and the relative ease for adults of learning to read text with and without sound-symbol clues to the orthography.  
Experiments 5.2 and 5.3 illustrated the relationship of reading and spelling, and the difficulty that even literate adults have with English spelling.

Demonstration Experiment 6.2 showed the impossibility of using Paulo Freire’s combined ‘top-down’ and ‘bottom-up’ pedagogy for learning to read with English spelling as it is.

Chapter 2 analysed the degree to which present English spelling operates by alphabetic, morphemic, or etymological principles, and concluded that English  homophones do not represent a real drawback to the alphabetic principle.

The abilities and needs of readers as determinants of  an optimum spelling.

Experiment 5.1 explored how readers’ strategies differ according to their experience of writing systems.  Study 6.6. showed the importance of individual differences in how and why children learn.

Experiment 7.1 showed the importance of the most familiar spelling for recognition of words which have alternative dictionary spellings, but Experiment 6.4 showed how readers apply untaught analogical orthographic principles for word recognition.  This is supported by features of the larger unpublished experiment of practice in reading in a modified spelling (Experiment 9.4), in which readers used CS principles in spelling preference and similar tasks before any improved skill in the speed or accuracy of actual reading habits was manifest. 

Spelling with surplus letters omitted (CS)

  Studies of English spelling structure and history (Chapters 3 and 7) indicated that  CS  Spelling principles may have some validity, and this was supported by Study 6.5 of young children’s own tendencies in spelling.   Experiments 6.3 and 8.1 suggest that CS could have advantages for children and foreign learners in identifying and using English vocabulary.

Findings of experiments in Chapter 9.

When CS spellings were encountered in text for the first time, readers adapted rapidly to become not significantly slower or less comprehending than when reading TO.  Words in isolation may at first be recognised less quickly, due to the word frequency effect, which Experiment 3.4 (dictionary spellings) showed applies to the orthographic appearance of words.   However, CS was shown to be not just a form of mutilation, but representing a basic word-structure used by readers.  When words were flashed briefly at 30 msec. in CRT experiments, rejections as non-words ran at approximately 10% of words in standard spelling (TO),  25% of CS words and 50% of RCS (random-cut control).   In a self-timed presentation of words, 1% of TO words were rejected, 12% of CS and again around 50% of control-cut words.   There was no significant difference in response latency between TO or CS for the 88% of cut-spelled words that were recognised as words, even though the spellings were novel to the subjects, but RCS responses were significantly slower.

CS spellings clearly represented significant word-structure for word recognition, but control-cut deletions mutilated this structure as often as not.  RCS rejections and response rates were significantly mitigated by pronounceability or orthographic legality.   Pronounceability (phonological factor) tended to be more relevant than orthographic illegality (visual factor).   Orthographic illegality affected recognition of CS to a lesser degree.  Findings replicated, and were reliable across subject-groups and across word sets. It is possible that findings for the self-timed presentations of words reflected inspection processes as much as or even rather than automatic processes, but inspection was hardly possible for the flashed Brief presentations.  It can also be argued that inspection plays an essential role in many if not most forms of reading.

Children who were good readers were similar to adult readers in their adaptation to CS.   Responses to  CS  spellings tended to be closest to TO  for the younger good readers, suggesting their greater initial reliance on phonological strategies.

All groups, Australian and ESL, adults and children, except for the poor readers, showed word recognition strategies that made some use of orthographic structure  - CS was significantly easier than RCS -  as well as of familiar visual configuration which facilitated TO.  

Weaker readers in the samples tended to be be equally slow and error-prone in all spelling modes (TO, CS and RCS) and pseudowords. They showed a limited range of vocabulary and lack of visual and phonological word recognition strategies.

Practice effects in word recognition require further reserch.  Pilot studies with only four subjects  found no effects with limited spaced repetition.  When two subjects undertook massed practice with 6 trials, with a total of 12 different word lists, all response latencies dropped significantly after the first trial, but did not change significantly from then on.   Scores originally tending to favour TO over CS converged at the third trial, tended to favour CS at the fourth, and began to show fatigue effects at the fifth.  However, high error rates and slow response latencies for RCS did not improve. 

Individuals differed in their trade-off between speed and accuracy, and slow responses in CS could be associated with accuracy, but overall the better readers were both fast and accurate.  Individual differences between subjects can be considerable and require investigating if full understanding of reading processes is not to be based on the blended mean. 

Differences in responses to particular modified words could  reach significance in larger studies, and could be investigated to refine the principles for optimum  CS  spellings. Trends suggested that word recognition was least affected and might be facilitated by deletion of doubled consonants that do not indicate phonology.   Final silent <e> tended to appear superfluous if serving no phonological function.  However,  CS  spellings were less immediately recognisable if two or more letters were deleted, or if short very common words were cut.

The significance of familiarity.   Orthographic frequency was a more important factor than simplicity of spelling in Experiment 5.1, which found that when dictionaries offered alternative spellings, the spelling given priority  was usually recognised more easily than the spelling given second place.   However, three subjects, familiar with cut spelling from participation in an earlier experiment, recognised the simpler spellings more quickly than the more complex ones, suggesting that familiarity with CS principles could develop quickly.   This was supported by Experiment 6.4, where there appeared to be transfer by analogy of  spelling patterns from a first word list to aid recognition of words with those same patterns in a second list.  This indicates that transfer of training for CS principles could occur rapidly, using analogical generalisation, and so not required by word memorising or calculation.

CS Spelling in semantic tasks.  In Experiment 9.3 the trend for semantic decisions to be made faster for word-pairs in TO was not significant, and a third of child and adult subjects read CS faster than TO.   There was no significant difference between comprehension responses in TO and CS.   Both were equally accurate, but CS tended to be 10% slower.  No practice effects were found after 36 sentences in cut spelling.

Orthography and the mental lexicon.  D. Bradley and M. Taft have conducted unpublished experiments in lexical decision with masked primes (Appendix  9.3 )with two different designs using CS word lists supplied by the writer. The aims were to test whether CS spelling structure of words might reflect how words are represented in the mental lexicon which is accessed for automatic word recognition, as indicated by their effectiveness as masked primes in lexical decision experiments.   No facilitation was found in either experiment, and several interpretations are possible.

Subjects were not found to show a CS-type orthographic structure for their mental lexicons, which may be closer to TO visual orthographic appearance, if masked primes are valid indicators.  Only one subject, a fluent reader with experience of CS, showed any effect - which suggests that practice effects may operate.  It is conceivable again, that  readers’ strategies may differ, and highly skilled readers may be found to differ from the average undergraduate.

For the present, however, it may be assumed that the hypothesis of a particular underlying orthographic structure of the reader’s mental lexicon is not supported.   It is possible that individual reading strategies of subjects are related to the form of successful primes for TO words, and I would like to see this experiment carried out with more investigation of the reading skills and strategies of the subjects taking part.

Silent reading of discourse

In a previous experiment with 90 subjects (Experiment 9.4) , CS reading rates and comprehension were also generally not significantly lower than for TO but no practice effects were found following in reading in CS over 7- 20 week-spaced one-hour sessions, except that poor readers in the CS group improved in their reading of TO text, and average readers made greater use of CS principles in a paper-and-pencil spelling preference tasks - as if the principles were transferring before any improvement in the visual-semantic reading task.    However, a replication is required with more rigorous timing than subjects’ self-timing with stopwatch, which was sometimes more erratic as the experiment progressed.   Oral reading in CS (which requires pronunciation assembly), while originally slower, improved in speed and accuracy to be not significantly different from TO, as soon as the second trial of reading, using parallel forms in a standard reading test.

CS  Spelling  in learning to read.  Children learning to read who had some understanding of sound-symbol relationships tended to read more words independently if they were in CS, even though the spelling mode was unfamiliar to them - suggesting that they might be further facilitated with more experience of it (Experiment 6.3).

ESL students, notably Chinese, probably having learned mainly from books in their home countries, tended to be more dependent on the visual forms of the word, and had more difficulties than Australian adults with novel CS word recognition, although their oral reading was improved (Experiment 8.1).   Experiment such as 5.3 and 9.5.1 showed that even long-term residents continued to have significantly more difficulties with TO spelling than Australian adults. The pilot study of Chinese students reading in different scripts (Experiment 5.1) suggested both the difficulty of an alphabetic representation of a multi-homophone tone language, which requires different strategies to English, and the significance of initial reading strategies for reading in a different orthography.

 CS  Spelling in writing and reading

In Experiment 5.3, subjects were asked to complete correctly the spellings of 16 common words presented in CS.  All except foreign students had no difficulty in recognising the CS words, but even experts in education and psychology could not replace all the deleted letters in the correct places, while secondary students averaged only 50% correct.  The implications are that the CS deleted letters serve no function in word identification - and particularly handicap spellers.

A pilot study , Experiment 9.5, rating the CS spellings of a list of 45 words on a 5 point scale, showed that subjects operate on principles of orthography that could be clarified by further study with this strategy.