7.1
Introduction
7.1.1 Unresolved outcomes
The following observations were made in the earlier experiments:
1. The Auslan and SE groups recalled the Lipsim list relatively poorly in Experiment 1. (This list contained items which made similar shapes "on the lips" when spoken, and which were also orthographically similar.)
2. In Experiment 3, the Auslan subjects were less sensitive to sign length variations than were the Oral group to English length manipulations.
3. The item length effects for all deaf subjects in Experiments 2 and 3 were smaller than those typically obtained with hearing subjects.
4. The imposition o f manual and English suppression in Experiments 4 and 5 failed to totally disrupt the memory performance o f the target groups.
All these results could be explained on the assumption that subjects were using an orthographic (or graphemic) code, based on the physical appearance o f the words and letters on the TV screen. This assumption is consistent with recent evidence from hearing subjects that visual short-term memory can support the retention o f several verbal items (Walker, Hitch, & Duroe, 1993).
In the first experiment, the two signing groups' unexpected sensitivity to the lipsim dimension could be understood in terms o f the greater visual similarity o f the items in that list, which would have detracted from their recall if they employed a visual code in addition to a cherological one. An analogous case could be made for the Oral group. (This point was elaborated in Section 4.5.1.)
Secondly, in Experiment 3, orthographic length and the duration o f spoken (but not signed) items correlated perfectly, hence visual coding by all subjects would have inflated the effect o f English (but not sign) length. As outlined in Section 5.5.5, this would explain how the Oral group's English length effect exceeded the Auslan group's sign length effect.
The third piece o f evidence favouring orthographic coding also came from Experiment 3. The finding that the item length effects observed for the deaf subjects were less than those reported for hearing people's recall implies that the deaf subjects were using strategies other than cherological and phonological rehearsal. (Sections 5.5.1 and 5.5.2 discussed this possibility.)
Finally, it was observed in Experiments 4 and 5 that neither manual nor articulatory suppression reduced the performance o f the Auslan and Oral groups respectively to chance. It was therefore suggested in Section 6.5.1 that these subjects were resorting to an additional non-cherological, non-phonological strategy to assist their memory.
This pattern o f results suggests two possibilities: First, that orthographic coding may have been occurring, and second, that this graphemic code was being used in conjunction with a communication-based code: it could not have been used merely as a replacement code when speech- or sign-rehearsal were prevented, otherwise outcomes such as the inflation o f the English length effect would not have happened.
Evidence that each o f these effects can occur will be reviewed in turn.
7.1.2 Graphemic coding by deaf subjects
Anecdotal support for visual coding can be found in the introspections o f deaf people themselves. Dimmock (1985), himself deaf, observed that:
normal people use internal speech as a means for self-communication. As for the deaf they use internal reading .... Most o f us are able to visualise whole sentences without relying on how they are sounded and to know the meaning instantly .... The mind itself is an efficient translator. Words become pictures and vice-versa (P- 5).
These assertions have been confirmed by empirical research. Graphemic coding has been reported by Budde and Jungnitsch (1983), Conrad (1970, 1971a), Locke and Locke (1971), MacDougall (1979), and Wallace and Corballis (1973). For example, many o f the errors in Conrad and Rush's (1965) data seemed to be visually based (e.g. R/B, Y/K, P/R, V/X, X/Y). Similarly, Ronnberg and Nilsson (1987) reported that hearing impaired subjects performed better than a matched control group in the recency portions o f the serial position curve for visually presented items, with the effect especially marked at slow presentation rates. They interpreted this to mean that the hearing-impaired subjects were coding the stimuli visually, therefore they did not have to recode the information for a written output mode.
D eaf subjects are influenced by orthographic similarity during rhyme judgements (Hanson & Fowler, 1987), during word association and recall (Blanton, Nunnally, & Odom, 1967), and during word recognition and spelling (Dodd, 1980; Doehring & Rosenstein, 1960; Frumkin & Anisfeld, 1977; Gibson, Shurcliff, & Yonas 1970; Hanson, 1982b, 1986; Hanson, Shankweiler, & Fischer, 1983; Stone, 1980). The latter is quite consistent with deaf people's relatively good spelling for the words they know (despite Rack and Snowling's (1988) observation that remembering via orthographic characteristics does not mean that the order o f letters is coded, therefore such a code does not automatically lead to good spelling).
Evidence o f an enhanced ability to use a sequential visual code in deaf subjects is provided by the findings that they performed better than their hearing counterparts on tasks involving the mental rotation o f images (Arnold, 1978) and the serial recall of photographs (O'Connor & Hermelin, 1973a). Direct evidence suggesting orthographic encoding is provided by the fact that Blair (1957) and Hermelin and O'Connor (1975)
found no difference between forward and backward digit span, suggesting that deaf children coded the material in a visual array which could be read equally well in either direction.
N ot all research with deaf subjects has supported the existence o f visual coding. For example, Hanson, Liberman, and Shankweiler (1984), and Hanson and Wilkenfeld (1985) concluded otherwise - but their subjects were atypical, highly proficient deaf readers, unlike the majority o f participants in the present study.
7.1.3 Multiple coding
In several o f the studies reviewed above there was some confounding between acoustic, articulatory, fingerspelling, and orthographic similarity o f items. Therefore, effects attributed to orthographic coding may have actually had some other basis. However, it has been established that graphemic similarity contributes independently to deaf subjects' performance. Wallace and Corballis (1973) and MacDougall (1979) used upper and lower case letters and different type scripts respectively to eliminate this overlap o f similarity dimensions. In both conditions the results supported the considerable use o f non-iconic visual coding in short-term memory tasks, in conjunction with other strategies. A literature review by Kettrick and Hatfield (1986) also observed that the coding spectrum used by deaf people to memorise English words and letters covered manual, visual, and phonological aspects. For example, Siple and Brewer (1985), investigating whether visual or motor tasks interfered with sign memory, found that some signers seemed to rely on speech, others on manual-sign and others on visual- sign.
Further indications that more than one code can be operative at the one time are provided by the results o f Hamilton and Holtzman (1989). These researchers found that subjects with both sign and speech experience recalled Signed English presentations better than ones presented via sign or speech alone.
7.1.4 Rationale for the experiment
In this experiment, it was critical to devise a test which would unambiguously diagnose the presence or absence o f visual coding. Such coding could conceivably be obscured by the sign- and speech-based strategies, which the present subjects had been shown to use in the other experiments. It was also necessary to determine whether visual coding could be used in combination with the cherological and phonological coding demonstrated in the earlier experiments.
In order to meet the discriminability requirement, a Vissim list o f visually similar items was designed together with a Control list in which the items were highly distinct. Impaired performance on the first list relative to the control list would indicate that the subjects were resorting to visual coding.
The second criterion - that the test should establish visual coding in conjunction with manual and/or phonological coding - was satisfied by using a multiple suppression procedure. The Vissim and Control lists were each presented three times - with no suppression, with manual suppression, and with English suppression. The logic underlying this methodology can be illustrated by referring to the Auslan subjects. If their reliance on visual coding was not affected by the presence o f manual coding, then the
degree o f impairment caused by visual similarity would remain constant irrespective o f whether manual coding were suppressed or otherwise. Similarly, if their reliance on manual rehearsal was not influenced by the presence o f visual coding, then the memory impairment caused by the elimination o f sign rehearsal would remain constant regardless o f whether visual coding were diminished or otherwise.
A similar argument could be made for the Oral group, for whom no interactions between visual similarity and English suppression would be expected.
7.1.5 Design overview
A repeated measures split-plot design was used, with three subject groups, two visual similarity conditions and three suppression conditions.
The subject groups were identical to those used in all previous experiments. These groups were:
Group 1, (Auslan), who employed sign only; Group 2, (SE), who used sign and speech; Group 3, (Oral), who communicated via speech.
The visual similarity conditions were called Vissim (high inter-item orthographic similarity) and Control (low inter-item orthographic similarity). These were implemented via two master lists o f eight items each. The design followed by all subjects is illustrated in Table 7.1.