Acta Psychologica 39 (1975), 91-96 0 North-Holland Publishing Company

INDIVIDUAL DIFFERENCES IN SPEED OF PHONEMIC ANALYSIS, VISUAL ANALYSIS, AND READING’

Jonathan BARON2 and Brian J. McKILLOP Dept. of Psychology, University of Pennsylvania, Philadelphia, Pa. 19174, U.S.A.

Received November 1974

College student subjects (Ss) classified phrases as making sense or not according to

sound in one condition, according to visual appearance in another, and in a basic task in

which either strategy would yield right answers. Ss who were relatively fast in the visual condition were faster at the basic task than Ss who were relatively fast in the sound

condition. Other results suggested that the former Ss used a visual strategy in the basic

task while the latter used a phonemic strategy.

Two different strategies are available to readers of English for iden- tifying a printed word. The phonemic strategy involves first translating the word into a full phonemic (auditory and/or articulatory) represen- tation, and then using this representation to retrieve the meaning of the word. This second step relies on the same knowledge used in identifying words in spoken language. This strategy must be used when we en- counter for the first time a word we have heard but not seen. The visual strategy involves using the visual information itself (or possibly some derivative from it which is not formally equivalent to overt pronuncia- tion) to retrieve the meaning. This strategy uses knowledge specific to reading. It must be used to distinguish homophones when the context is insufficient, for example, in the sentence; ‘Give me a pair (pear).’

Baron (1973) has argued that a visual analysis is faster for normal

r This work was supported by a grant from the National Research Council of Canada. It is based on a B.A. thesis by the second author, supervised by the first, in 1973. The thesis

contains the data for all Ss and lists of the stimuli. It is on file in the Psychology Department,

McMaster University, Hamilton, Ontario, Canada, L8S 4K1, where it was done.

* Reprints may be obtained from J. Baron, Department of Psychology, University of Pennsyl- vania, Philadelphia, Pa. 19174.

92 J. Baron, B.J.McKillop/Phonemic and visual analysis and reading

readers, and is thus normally used when the reader is free to choose his strategy of analysis. Baron asked college student subjects (Ss) to classify phrases as sense or nonsense. Two kinds of phrases were considered to be nonsense, phrases that were true nonsense (I AM KILL), and those which sounded sensible because of a homophone (ITS KNOT SO), which we shall call ‘homophone phrases’. Ss took no longer to classify the homophone phrases as nonsense than the true nonsense phrases, in spite of the fact that their phonemic representation made sense. Thus, phonemic conflict had little effect on this task. In a second experiment, Ss were asked to consider the homophone phrases as sense, that is, to use a phonemic analysis for these stimuli. In this case, true sense (ITS NOT SO) was classified more quickly than the homophone phrases, indicating that visual information was available early in the analysis of each phrase.

If, as suggested by these findings, the visual strategy is normally quicker, we would expect the speed of visual analysis to be a better predictor of an individual’s reading speed than his speed of phonemic analysis would be. The present study attempts to test this expectation by examining individual differences in phonemic analysis, visual analy- sis, and a basic reading task in which Ss are free to use either strategy. Lists were constructed of 12 four-word phrases, each of which could be true sense (S), such as ‘a victim of circumstances’,. nonsense (N), such as ‘his temperature remains contest’, or homophone phrases (H), as de- scribed above, such as ‘a bat and bawl’ (see fn. 1). Each list contained two different kinds of items. Thus there were three kinds of lists, which we shall call SN, HN, or SH. Ss were asked to categorize the phrases as quickly as possible into the two types, with a ‘yes’ response used for S in SN, H in HN, and S in SH lists. It was assumed that a phonemic strategy would be used for the HN lists, that is, that the S would distinguish H phrases from N phrases on the basis of whether their sound made sense; a visual analysis would render all phrases nonsense. Likewise, it was assumed that a visual strategy would be used for the SH lists; if a phonemic strategy were used, all phrases would seem sensible. The S was free to use either strategy on the SN lists. According to our hypothesis, however, at least the fast readers would use a visual strategy on the SN lists. Thus, it was predicted that an S’s speed on SH lists would be a better predictor of his speed on SN lists than would his speed on HN lists.

While the task of classifying a phrase as sense or nonsense may not be

J. Baron, B.J. McKillop/Phonemic and visual analysis and reading 93

a pertect model of real reading, it does require at least some semantic analysis. Further, we suspect that judgments of sense actually make up a considerable part of some types of reading. In studying, for example, a useful strategy may be to decide whether each new sentence ‘makes

sense’ in that it fits in with what we already know, and if not, to analyse it more carefully.

Method

There were two parts to the experiment. The first part used SN and SH lists, the second

part, SN and HN lists. Each list contained 12 phrases, and each part used 5 lists of each type. In

part 1, the SN lists used 30 phrases of each type, and all other lists likewise used equal numbers

of each phrase type within a part. The order of the two types of phrases within a list was

pseudo-random. Phrases were typed triple-spaced, 12 to a page.

In part 1, Ss were told to read down each list and say ‘yes’ if a phrase made sense ‘no’ if it

did not, as fast as possible without errors. They were told not to waste time trying to make

sense out of nonsense phrases, and that the difference would be obvious. Examples of each type

of phrase were given. The Experimenter tried to make sure by repeated explanation that the S

understood how the lists were made up. Each list was presented once to each S. The two kinds

of lists were alternated. Lists were presented in the same order to all Ss. Before each list, S was

told which type it would be. This was to allow him to choose his preferred strategy for SN lists,

visual or phonemic, and to inform him about the strategy required for SH and HN lists.

Instructions and procedure for part 2 were the same, except that Ss were told to respond

according to whether or not the phrases sounded as if they made sense. As in part 1, it was

made clear to the Ss that they could use either strategy on the SN lists. There was a brief rest period between the two parts.

Forty Ss were run. They were students (graduate and undergraduate) of McMaster Univer-

sity, of both sexes, and were all native English speakers. They were paid for participation.

Results and Discussion

The basic data were the times required by each S for each of the three kinds of list. Data from SN lists from parts 1 and 2 were com- bined. To test our hypothesis that SH time would be a better predictor of SN time than would HN time, the ratio of SH time to HN time was computed for each S (except those who made 10 or more errors on any of the three kinds of list). It is assumed that those Ss with a low SH/HN ratio are relatively good at the visual strategy, compared to other Ss, and that those with a high SH/HN ratio are relatively good at the phonemic strategy. According to our hypothesis, the visual Ss, those with a low SH/HN ratio, ought to be faster on the SN lists, i.e., those lists on which either strategy could be used.

94 J. Baron, B.J. McKillopjPhonemic and visual analysis and reading

To test our hypothesis, two groups of Ss were selected for compari- son, the 5 Ss with the lowest SH/HN ratio, which we shall call the ‘visual group’, and the 5 Ss with the highest ratio, the ‘phonemic group’. (As shown in table 1, error rates do not contradict times as measures of difficulty.) The reason for using such small groups was to take advantage of the possibility that Ss who use predominantly one strategy or the other on SN lists might be rare. In particular, from the results of Baron (1973) we would expect few Ss to use a phonemic strategy on SN lists. The use of small groups makes it likely that such Ss, if present, would contribute substantially to our results.

Data for the two groups are shown in table 1. According to our hypothesis, the visual group should be faster on the SN lists. This was the case 0, < 0.02, Mann-Whitney U-test, two-tailed). Thus, Ss who were relatively good at the visual strategy were faster when they were free to use either strategy. (The correlation coefficient over the entire sample, excluding Ss who made too many errors as before, was 0.20, which was not significant; we ascribe this to the rarity of ‘phonemic Ss’, as explained above.)

There are two explanations of this relation between SH/HN ratio and free (i.e., SN) speed which seem worthy of further consideration. The first explanation assumes that the fastest strategy determines the re- sponse on the SN lists, but this strategy tends to differ for different Ss. The phonemic group would thus be more likely to use a phonemic strategy on a given phrase. The second explanation is that all Ss use a visual strategy almost exclusively on SN lists. Ss who are better at this strategy are thus faster on both SN and SH lists, but not necessarily any faster on HN lists, which require a different strategy.

If the second explanation were correct, SN and SH times should

Table 1

Summary of data for groups high or low in SH/HN ratio.

Mean set per list Mean % errors

Group SN SH HN SN SH HN

Low ratio

(visual Ss)

High ratio

(phonemic Ss)

15.7 15.1 21.1 3.2 4.3 7.3

20.5 24.8 21.6 3.0 1.3 8.3

J. Baron, B.J. McKillop/Phonemic and visual analysis and reading 95

serve as independent estimates of the speed of the visual strategy. Thus we ought to get a better estimate of the relative speed of the two strategies by using the mean of the SN* and SH* times instead of the SH time alone, where SN*, for example, is the normalized score (SN divided by the average SN across Ss), which is used to ensure that the contribution each score makes to the mean of the two does not depend on uncontrolled stimulus variables which affect time. Based on this assumption, two new groups of 5 Ss each were selected (again excluding

Ss with 10 or more errors on any type of list), with the highest and lowest (SN* + SH*)/2HN* ratios. If our assumption is correct that SN and SH times are independent estimates of the speed of the visual strategy, then the ratio SH/SN ought to be constant for these two new groups. This was not the case. The mean SH/SN ratio for the ‘new visual’ group (low (SN* + SH*)/2HN* ratio) was 0.984, and for the ‘new phonemic’ group (high ratio), 1 .192 (p < 0.02, U-test, one-tailed). It would thus appear that SN times are not simply estimates of visual strategy speed, and that SN times are relatively faster than SH times for ‘phonemic’ Ss because for these Ss the phonemic strategy is often faster than the visual strategy.

The existence of predictable individual differences in the relative times required for SH and SN lists rules out a full interpretation of our results in terms of any properties of the sense items alone, such as differential sensitivity of Ss to their familiarity (Meyer et al. 1974: pp. 310-3 11) or to the appearance of easiness that they present.

It thus seems from our results that readers who are better at the phonemic strategy than the visual strategy are slower than readers who are better at the visual strategy, in a population made up of experienced and competent readers. We may thus conclude, in agreement with Baron (1973), that the visual strategy is a generally superior one. We may also conclude, in agreement with Hardyck and Petrinovich (1970), that some readers persist in using a phonemic strategy, in spite of the fact that a visual strategy might be faster. This finding appears to con- flict with Baron’s (1973) claim that the phonemic strategy was rarely used by college Ss. It is our impression, however, that the use of the phonemic strategy (as indicated by a high SH/SN ratio) was limited to a small proportion of Ss. Baron’s (1973) sample was much smaller than ours and could easily have excluded all Ss of this type.

The interesting question remaining is why some Ss persist in using a slower strategy. One possibility is that they simply have not learned to

96 J. Baron, B.J. McKillopfPhonemic and visual analysis and reading

use the visual strategy efficiently. Use of the visual strategy does not allow the reader to rely as heavily on his knowledge of the relation between spoken language and meaning, or on his knowledge of the relation between spelling and sound. Thus the use of the visual strategy requires knowledge which is redundant with other knowledge which must be possessed anyway. This redundant knowledge is that of the direct relation between spelling (or visual form) and meaning. The Ss who are better at the phonemic strategy may not have practiced using this knowledge enough to do it efficiently. (Practice at writing may not transfer well to reading, since the direction of transmission is reversed.) If this explanation is correct, it should be possible to improve the reading speed of the ‘phonemic readers’ by giving them extensive prac- tice at using the visual strategy. Methods useful for this might be sup- pressing articulation (Hardyck and Petrinovich 1970), forcing irrelevant articulation, or using our SH condition. The method of testing used here might be useful, in this case, for finding out which readers could be helped by such practice.

The other reason why some Ss may use the phonemic strategy is that this strategy might be more efficient for all readers under certain condi- tions. For example, the phonemic strategy may be advantageous when the material is ‘difficult’ in some sense (Hardyck and Petrinovich 1970), perhaps because a phonemic code is easier to retain in memory while extraction of meaning proceeds. Thus the crucial differences between our groups might not be their relative skills at phonemic and visual strategies per se, but rather their skill at some other process common to both strategies, such as construction or synthesis of meaning. If this were the case, practice at the visual strategy would be no more useful than any other sort of practice involving the process which creates the difficulty.

References

Baron, J., 1973. Phonemic stage not necessary for reading. Quart. J. Exp. Psychol. 25, 241- 246.

Hardyck, C.D., L.F. Petrinovich, 1970. Subvocal speech and comprehension level as a function

of the difficulty level of reading material. J. Verb. Learning Verb. Behav. 9,647-652.

Meyer, D.E., R.W. Schvaneveldt, M.G. Ruddy, 1974. Functions of graphemic and phonemic codes in visual word-recognition. Memory and Cognition 2,309-321.