Assessing the benefits of phonics intervention on hearing impaired children's word reading

Sue Palmer

University of Manchester

Paper presented at the British Educational Research Association Conference, Cardiff University, 7-10 September 2000.

 

Reference:- Deafness & Education International 2(3), 2000 pp165-178

Address for Correspondence
Dr Sue Palmer
Education Support and Inclusion
School of Education
University of Manchester
Oxford Rd
Manchester
M13 9PL
Email s.palmer@man.ac.uk
Tel 0161 275 3381
Fax 0161 275 3373

ABSTRACT

Previous studies demonstrate that phoneme awareness training, particularly when combined with grapheme-phoneme correspondence (letter to sound matching) teaching, results in improved reading and spelling development for normally hearing children. This study seeks to investigate the efficacy of these claims for hearing impaired children. Two hearing impaired children were pre-tested on measures of spoken and written language, phonological abilities and alphabet knowledge prior to a 12 week intervention using a Phonographix teaching programme. The children were again tested on all measures immediately after intervention. The intervention programme accelerated the children's acquisition of phoneme awareness and of phoneme-grapheme correspondence knowledge, and their ability to apply these in reading and spelling. Can concentration on teaching phoneme awareness and phonics radically improve reading and spelling standards of hearing impaired children despite a poor standard of receptive language?

INTRODUCTION

Written English language is a phonetic code. Each phoneme in a spoken word is represented by a grapheme in a written word. Although numerous studies have shown that early phonological skills training has beneficial effects on reading development (e.g. Cunningham, 1990; Duncan and Johnston, 1999), other training studies have shown that phonological awareness on its own, although necessary, is not sufficient. Phonological skills training and phoneme-grapheme correspondence instruction are both necessary components for maximum benefit (Berninger, Abbot, Zook, Ogier, Lemos-Britten and Brooksher, 1999; Bus and van Ijzendoorn, 1999). Malicky and Norman (1999) have suggested that, rather than phonemic awareness per se, what seems to be essential to learning to read is that children develop an understanding of the connections between oral and written language. At the macro level this involves an understanding that written words represent words in oral language. At the micro level it involves understanding that letters, or groups of letters, in written words stand for individual sounds in spoken words.

The majority of new readers come to school with neither the skill to hear the individual sounds in words nor the knowledge that written words represent spoken words and even more rarely with the knowledge that the letters in the written word represent the sounds in the spoken word. Those who do not receive explicit instruction in the link between spoken and written language can begin to make false assumptions about written language. Many children believe that they must remember the whole word, rather than decode the word one sound at a time. They do not understand the fundamental intention of written English language. Once children begin to memorise whole words as a primary reading strategy, they have developed a bad habit which will make learning to read and spell correctly more difficult.

Since 1998, the National Literacy Strategy (NLS) Framework, has required schools and teachers to incorporate phonological skills training and phoneme-grapheme correspondence into the early reading curriculum. The phonic work in the NLS Framework is built around the idea that spoken English language is encoded in approximately 44 phonemes and represented by 26 letters in about 140 graphemes in written English language. Children are required to learn to identify the phonemes in their spoken language and learn how these phonemes are commonly spelt, how they can be blended, segmented and manipulated.

They need to understand that: some phonemes are represented by graphemes consisting of one letter, others have more than one letter; many phonemes can be represented by more than one grapheme, whilst some graphemes can represent more than one phoneme. Thus children need to develop four basic skills to promote and develop independent writing and fluent reading. They must develop the ability to:

Identify sounds in spoken words (phonological awareness)

Recognise the common spellings for each phoneme (phoneme-grapheme correspondence)

Blend phonemes into words for reading

Segment words into phonemes for spelling.

In addition, children will need to develop a fifth skill, which will allow them to manipulate the phonemes in different parts of words for reading by analogy, a skill that will allow them to decode new words unaided.

The NLS framework is intended for use by all teachers with all children, regardless of their sensory status. Children with hearing impairments are at an immediate disadvantage when learning to read on two main counts.

Hearing children usually enter school with a broad knowledge of words and their meanings. We cannot make such assumptions for hearing impaired children. Reading is dependent upon, and reflects children's linguistic, cognitive and experiential development. Even where the hearing impaired child has been in a pre-school environment where books have been in abundance and whose cognitive development is on a par with his or her hearing peers, the overarching problem for hearing impaired children learning to read is one of language.

In addition, even when they have had access to appropriate, early and consistent amplification, they may not have easy access to all the sounds within words and hence may develop inaccurate phonological representations, if indeed they develop them at all. Children educated in an entirely signing environment might develop the connection between 'spoken' and written words at a macro level, but often not at a micro level apart from fingerspelling. While children educated within auditory/oral environments can achieve micro level correspondence through amplified residual hearing, confusions may still arise when sounds are outside the child's hearing range, or are confusable in lip shape.

Leybaert (1998) has argued that the development of phonological representations in deaf children does not necessarily depend on auditory experience, neither at the perception nor at the production level. Instead, this development depends upon early experience of an input in which all phonological contrasts are well specified, independent of input modality. Leybaert and Alegria (1995) found that in spelling, as well as reading, hearing impaired children rely on inaccurate speech representations derived mainly from lip-reading and hence are more likely to show errors of confusion consistent with lip-readability.

The benefit of early experience of well specified phonological contrasts and consistent amplification is dependent on early diagnosis, early fitting of appropriate hearing aids and access to a good pre-school teacher of the deaf using appropriate teaching methods which are conducive to later literacy acquisition.

Prior to 1998, few teachers of the deaf of school-age children used phonics teaching with hearing impaired children. Many teachers employed methods where the child learned the word as a whole unit. This strategy is fine up to a point but it does mean that each word has to be learned individually with no facility for self teaching or reading by analogy. Consequently only a small minority of children with pre-lingual onset, severe to profound hearing impairments acquire high literacy skills (Nelson and Camarata, 1996).

THE STUDY

This study set out to evaluate research findings of the beneficial effects of phonological skills and phoneme-grapheme correspondence training in relation to children with severe to profound hearing impairments. The researcher was a trained Phonographix trainer (see McGuinness and McGuinness, 1998 for a full description of research behind and methods used in the Phonographix programme). Phonographix teaching methods are compatible with the NLS framework for the teaching of phonics but takes a more structured approach. It is based on five basic premises

  1. written English is a coded form of spoken English
  2. spoken words consist of sounds that can be represented by word pictures (graphemes)
  3. word pictures can consist of one or more letters
  4. a word picture can represent more than one sound
  5. a sound can be represented by more than one word picture

and utilises the three basic phonics skills

  1. segmenting
  2. blending
  3. phoneme manipulation.

Phonographix uses real reading books right from the first lessons and sets learning of phoneme-grapheme correspondence and phonological skills in the context of reading books. The structure of this reading scheme was thought to be beneficial to teaching hearing impaired children.

The Phonographix programme was designed to be used by parents and teachers alike. Remedially, it is aimed at a twelve-week programme. This study was carried out over the same time period. The teacher of the deaf was given basic training in Phonographix methods and was provided with supplementary exercises to complete with the children on the days that the researcher was not in the school.

The following questions were investigated: 1) is it possible to teach hearing impaired children grapheme-phoneme correspondences rather than just letter names or finger spelling signs?; 2) is it possible to train hearing impaired children in phoneme blending, segmenting and manipulation skills?; 3) if so, does the acquisition of the these skills lead to more successful development of word decoding and spelling skills?; 4) if so, can these improved decoding skills be used to increase a child's vocabulary?

Participants

Two children, child A (female) aged 8 years 10 months and child B (male) 9years 0 months were referred by their unit teacher because they were underachieving in reading. The children attended a main stream primary school which had a hearing impaired unit staffed by a trained teacher of the deaf and two part-time assistants. Both children were pre-lingually deaf with profound bilateral sensori-neural hearing loss. Child A had unaided thresholds of 80dB at 250Hz down to 120dB at 4000Hz. Her aided thresholds were 30dB at 500 Hz down to 60dB at 4000 Hz. At the beginning of the study she appeared not to be using her residual hearing to any effect. Child B had unaided thresholds of 85dB in the right ear and 100dB in the left ear at 250 Hz down to 120dB in both ears at 4000 Hz. His aided thresholds were 35dB at 500 Hz down to 60dB at 2000 Hz with no response at 4000 Hz. Child A was a good user of her hearing aids whilst child B often forgot to bring them to school and was thought to rarely use them at home.

Child A was deafened through meningitis at 12 months and received hearing aids at 15 months. Her parents tried to communicate through lip reading but without much success. Neither parent is competent in sign. Child B was born deaf although he was not diagnosed until 18 months and received hearing aids at 2 years. His mother learned sign language and tried to teach him to communicate through sign.

Both children had attended pre-school classes with a teacher of the deaf. The hearing impaired unit at the primary school used a system of sign supported English. At the beginning of the study neither child had very intelligible speech although child A was slightly easier to understand then child B. Both children had limited knowledge of sign although they could make themselves understood by a mixture of signing and voicing. Both children were fairly good at lip-reading.

Teacher training in the use of Phonographix

The teacher of the deaf was given a copy of Reading Reflex (McGuinness and McGuinness, 1998) to read, which describes the programme, its rationale, and the materials provided to support teaching. She was also given a short training programme on the use of the materials. She was supplied with photocopiable worksheets to use with the children during the literacy hour phonics sessions.

Intervention

The intervention period was 12 weeks. The researcher spent one hour twice a week with the children in addition to the time they spent with the teacher of the deaf during the daily literacy hour. During this hour, the researcher gave individual attention to the children. Specific sounds were concentrated on during the sessions. For each group of sounds exercises were carried out involving segmenting, blending and phoneme manipulation using real words, non-words and simple books.

For the first twelve sessions, phoneme-grapheme correspondence was explicitly taught, including multiple correspondence. During this time the children progressed from simple cvc words to cvcc and ccvc words. The second half of the intervention period was spent consolidating the sounds learned and introducing more complex words include some multisyllabic words.

Testing

Prior to intervention the children were pre-tested on eight experimental measures of working memory encoding (Palmer, 2000), segmenting, blending, phoneme manipulation, letter-sound knowledge, Wide Range Achievement Test (WRAT) (reading and spelling) (Jastack and Wilkinson, 1984), nonword reading.

In addition five control measures were taken working memory span (auditory and visual), letter-name knowledge, Wide Range Achievement Test (WRAT) (maths) (Jastack and Wilkinson, 1984), British Picture Vocabulary Scales (BPVS), (Dunn, Dunn, Whetton and Burley, 1997), Raven's Coloured Progressive Matrices (Raven, 1996).

CONTROL MEASURES

Working memory span

The two versions of working memory span were included as a measure of the memory load with which the children could cope in different presentation modalities.

a) auditory span: the children were given a digit span test in which they were allowed to either sign or say the numbers given to them. The numbers were presented simultaneously in sign and word. The span was measured at the list length at which they could accurately repeat two out of three lists. A half credit was given if they repeated just one list of the appropriate length.

b) Visual span: using Corsi blocks, the children were required to replicate a sequence in which the blocks were touched in a random order. The span was measured as the sequence length at which they could accurately repeat two out of three sequences. A half credit was given if they repeated just one sequence of the appropriate length.

Letter name knowledge

This measure was included, as several studies have shown that knowledge of letter names can be advantageous in learning to read. The children's teacher had spent a lot of time teaching the children the names and/or signs of the letters but had not spent time on letter-sound correspondence.

Mathematics knowledge

The WRAT maths test was included to test the specificity of the intervention effects. If these are specific to literacy skills, then at post test children should differ in literacy skills but not in maths knowledge. If at post-test children also differ in maths knowledge, then some non-specific 'Hawthorne effect' may be operative.

Vocabulary knowledge

As vocabulary knowledge obviously influence literacy outcomes, it was important that some measure of this was taken. The children's receptive vocabulary was measured using the British Picture Vocabulary Scales (BPVS), (Dunn, Dunn, Whetton and Burley, 1997) which gives norms for children learning English as an additional language as well as for children whose first language is English.

Raven's Coloured Matrices

This test was included as a measure of non-verbal intelligence and was administered in accordance with written instructions. This test requires children to chose one of four alternative patterns with which to complete a design. It is arranged in three sets of 12 designs, each set becoming progressively more complex.

EXPERIMENTAL MEASURES

Working memory encoding

This measure was thought to be an important one to include. In normally hearing children, the degree of completeness phonological recoding, as shown by the relative size of the phonological and visual similarity effects, is a good predictor of reading level. Children who's central executive is unable to inhibit a visual response in preference to a phonological response are poorer readers than those in which this ability has developed (Palmer, in press). Teenage dyslexics were found to retain evidence of visual encoding long after their reading age match peers had stopped using it (Palmer, 2000). With hearing impaired children there is a third area of encoding which may be another potential source of confusion. Wilson and Emmory (1997) have shown that there can be a sign similarity effect in children who communicate through a signing medium.

The children were tested using four sets of pictures in which visual, phonological and sign similarity were manipulated. The pictures were presented in a serial recall test in which the span was measured for each set of pictures. As before span was considered to be the list length at which the child could recall two out of three lists with a half credit being given for one correct set recalled. The sets of pictures comprised visually similar set - ball, belt, cake, face, plate, ring, sun, wheel; phonologically similar set - bat, cap, cat, hat, fat, map, rat, tap; sign similar set - box, bridge, car, dog, house, pram, road, tent; a control set consisting of bed, chair, cup, drum, duck, fish, shoe, tree.

Phonological skills

The children were tested on the following measures

a) Segmenting - a phoneme segmentation task using 6 cvc words, 6c©v©c words and 6 non words was carried out after 5 practice items e.g. cat is /k/./ae/../t/. In total there were 63 phonemes to be segmented

b) Blending - a phoneme blending task using 9 cvc words and 6 c©v©c words with 5 practice items e.g. /k/./ae/../t/ is cat.

c) Auditory processing - a phoneme deletion task using 3 cvc words and 6 c©v©c where either the initial or final phoneme is deleted. E.g. say 'sip' without the 's'

Letter-sound knowledge

This was tested in two ways. Asking children to point to the letters that represent each of the 26 single letter phonemes spoken by the tester tested letter-sound recognition. The letters were presented in random order with 9 letters per page and letters that can represent the same sound e.g. 'k' and 'c' appearing on different pages. Showing children a card with 43 graphemes (single letters and digraphs) that represent 43 phonemes in English tested letter-sound recall. The researcher pointed to each grapheme in turn and asked the child to say its sound.

Reading and Spelling measures

The Wide Range Achievement reading and spelling subtests were used to measure the children's reading and spelling level (Jastack and Wilkinson, 1984). To avoid undue stress at pre-test, only the first ten words in both reading and spelling were presented. At post-test the child was allowed to carry on until 6 consecutive errors were made.

In addition to the word reading and spelling, the children were given a nonword reading test to test the degree to which they were phonologically decoding written symbols.

RESULTS

For control measures, there should be no appreciable gains or losses. This indeed proves to be the case (see table 1). If the intervention has been effective, then there should be changes in the scores of the experimental measures taking them closer to the age norm.

Table 1 about here

Control measures

Working memory span

a) Auditory - both children were initially below the norm for their peers. However, over the twelve-week period, they both gained very slightly. This may be an artefact of repeating the test or may be a reflection of improved auditory working memory span.

b) Visual - both children's visual span was above the mean for their age. This reflects the fact that both children work mainly through the medium of sight and therefore their visual ability may be expected to be better developed than that of their peers.

Letter-name knowledge

Both children were familiar with the names of all the letters of the alphabet. However, their knowledge consisted mainly of finger spelling letter symbols rather than oral names for many of the letter names. Child A knew the oral names for 20 of the 26 letters whilst child B knew only 15 oral names.

Maths knowledge

As table 1 shows, both children were performing at age on the maths test. There was no significant improvement in score on the maths test between pre and post testing showing that the intervention was specific to literacy.

Vocabulary knowledge

Both children's vocabulary scores are substantially below the standard score for their age. This is not unusual in hearing impaired children and may be a contributory factor in the poor levels of reading shown initially. The gains shown at post-test might reflect the fact that the teacher of the deaf began to ask the children to tell her which words they were reading they did not understand. In this way new vocabulary was learned as reading standard improved.

Raven's Progressive Coloured Matrices

The score for both children was close to or at the mean value for their age. This again is a common finding. Many hearing impaired children have normal non-verbal IQs but below average verbal IQs.

Experimental measures

Working memory encoding

As can be seen from table 1, both children were showing a phonological similarity effect in the same way as their peers would. However, initially, both children were also showing both a visual similarity effect and a small sign similarity effect. Their peers would demonstrate neither of these effects. At post-test, child A was no longer dependent on using sign encoding to recall a series of pictures although child B was still showing some signs of using sign to encode. However, both children were retaining visual encoding. The retention of the visual code may still be an obstacle to full phonological recoding and hence the ability to move completely from the logographic phase to the alphabetic phase of reading.

Phonological skills

Segmenting - from being unable to segment even the simplest words at pre-test, both children were able to segment more than half the phonemes. Errors were made in the main by segmenting into onset and rhyme rather than into individual phonemes e.g. pin was given as /p/ /in/. With a longer training period these errors could be overcome as when the errors were pointed out in a follow up session, child A was very quick to see where she had gone wrong. Child B did not always pick up on the final phonemes when the word was spoken to him but when a picture and the written word accompanied the word, he was then able to complete the task correctly. Segmenting is a skill necessary for accurate spelling from dictation. The teacher of the deaf was asked to give both children practice in dictation exercises as a follow up to the study.

Blending - this is a skill that is necessary for good word decoding. At pre-test child A could blend 4 of the simplest cvc words whilst child B could only blend 1. At post-test both children were able to blend all the cvc words but had not yet progressed to being able to cope with consonant blends. With further training, this skill may develop with oral presentation, as both children were able to read unfamiliar c©v©c words by the end of the study.

Phoneme deletion

At the beginning of the study, neither child could complete any of this task. The concept of deletion was not unfamiliar to them as they understood that if you took away half a compound word you would still be left with a word e.g. wheelchair take away wheel leaves chair. At post-test both children were performing at just under their age norms. During the intervention the children played word-chain games whereby they had to change the initial, medial or final phoneme of a cvc word e.g. make 'pot' into 'pat'. Make 'pat' into 'cat'. Make 'cat' into 'cap' etc. Although they were given no specific training in phoneme deletion, the phoneme manipulation of the word chain games had a carry-over effect, which allowed them to change one word into another.

Letter-sound knowledge

At pre-test neither child knew many letter-sounds. They both knew the sounds of the vowels and the sounds of some of the letters in their name. By post-testing, both children were well on the way to recognising and recalling the majority of sounds tested.

Word decoding

Although neither child was reading at their chronological age, both had made tremendous gains. They were able to decode the majority of regular words, even some multisyllabic words. Their errors were mainly irregular or less frequent words. Child A was now reading within one standard deviation of the age mean whilst child B was not far behind. With more context both children's scores might improve. However, when account is taken of their poor vocabulary score, their reading score is good. The children were given the opportunity to learn new vocabulary by capitalising on the improved decoding ability in combination with labelled pictures from different categories of words eg fruits, vegetables. The teacher of the deaf was going to continue to use the newfound decoding skill to improve the vocabulary knowledge.

Spelling

As with reading, the improvement in their segmenting ability is reflected in their increased spelling scores. Child A is now just a little more than one standard deviation below the norm whilst child B, although he showed the most dramatic improvements, still has a long way to go to catch up with his peers. However, the spelling test used moves rapidly from easy cvc and familiar sight words, to more complex words.

Non-word reading

Probably the most convincing evidence that both children were using some phonological recoding in their reading, is the fact that they were able to read some nonwords at post-test. Reading nonwords involves recognising the grapheme, recoding it into its phoneme and then blending the phonemes to make the whole word. This skill tests not only knowledge but also memory to hold sounds for blending.

DISCUSSION

The results presented show strong, positive effects of the Phonographix intervention on the development of grapheme-phoneme correspondence knowledge, and the acquisition of blending, segmenting and phoneme manipulation skills all of which lead to the development of word decoding and spelling skills.

This study provides support for the view that structured and focused teaching of phoneme segmenting, blending and manipulation skills and of grapheme-phoneme correspondence does accelerate development of these skills and acquisition of this knowledge in hearing impaired children. This accelerated development does lead to improved word decoding and spelling skills despite poor receptive vocabulary scores and indeed can be used to help develop vocabulary.

Both children in the study were in year 4 in school before the intervention was started, perhaps if intervention was started at the earliest stage of education, it might have even more beneficial effects, including improved comprehension skills. The National Literacy Strategy Framework has guidelines, which suggests that children are trained in segmenting and blending skills and taught grapheme-phoneme correspondences as pre-reading skills. This strategy needs to be implemented at least as early as the reception class for all hearing impaired children irrespective of the preferred mode of communication.

As an unexpected benefit of this intervention study, the teacher of the deaf reported that the behaviour of both children had improved. Their attention span had increased and they were more attentive to the oral aspects of lessons. In addition, child A's speech therapist reported that her speech had improved and that she was paying more attention to the sounds that she was being taught.

Acknowledgements

Thanks are due to the children and staff who took part in this study

REFERENCES

Ball, EW. and Blachman, BA. Does phoneme awareness training in kindergarten make a difference in early word recognition and developmental spelling? Reading Research Quarterly 1991; 26, 49-66

Berninger, VW., Abbott, RD., Zook, D., Ogier, S., Lemos-Britton, Z. and Brooksher, R. Early intervention for reading disabilities: Teaching the alphabet principle in a connectionist framework. Journal of Learning Disabilities 1999; 32, 491-503

Bus, AG. and van Ijzendoorn, MH. Phonological awareness and early reading: A meta-analysis of experimental training studies. Journal of Educational Psychology 1999; 91, 403-414

Cunningham, AE. Explicit versus implicit instruction in phonemic awareness. Journal of Experimental Child Psychology 1990; 50, 429-449

Dunn, LM., Dunn, LM., Whetton, C. and Burley, J. The British Picture Vocabulary Scales. 2nd edition. Windsor: NFER Nelson 1998

Duncan, LG. and Johnston, RS. How does phonological awareness relate to nonword reading skill amongst poor readers? Reading and Writing 1999; 11, 405-439

Jastak, JJ. And Wilkinson, G. Manual: Wide Range Achievement Test- Revised. Wilmington, DE: Jastak Associates 1984

Leybaert, J. Phonological representations in deaf children: the importance of linguistic experience. Scandinavian Journal of Psychology 1998; 39, 169-173

Leybaert, J. and Alegria, J. Spelling development in deaf and hearing children - evidence for use of morpho-phonological regularities in French. Reading and Writing 1995; 7, 89-109

McGuinness, C. and McGuinness, G. Reading Reflex. London: Penguin 1998

Malicky, GV. And Norman, CA. Phonological awareness and reading: An alternative interpretation of the literature from a clinical perspective. Alberta Journal of Educational Research 1999; 45, 18-34

Nelson, KE. and Camarata, SM. Improving English literacy and speech acquisition learning conditions for children with severe to profound hearing impairments. Volta review 1996; 98, 17-41

Palmer, SE. Phonological recoding deficit in working memory of dyslexic teenagers. Journal of Research in Reading 2000; 23, 28-40

Palmer, SE. Development of phonological recoding and literacy acquisition: A four-year cross-sequential study. British Journal of Developmental Psychology (in press);

Raven, JC. Raven's Coloured Progressive Matrices. Windsor: NFER Nelson 1996

Share, DL. Phonological recoding and self-teaching: Sine qua non of reading acquisition. Cognition 1995; 55, 353-366

The National Literacy Strategy: Framework for Teaching 1998; DFEE Standards and Effectiveness Unit

Wilson, M and Emmory, K A visuo-spatial 'phonological loop' in working memory: evidence from American Sign Language. Memory and Cognition 1997; 25, 313-320

 

Table 1 pre and post test results with age norms for comparison

Test

Child A

pre-test

Child A

post-test

Child B

pre-test

Child B

post-test

Age norms

Control measures

Auditory span

4

5

3

4

6

Visual span

6

6

8

8

5

Letter-name (26)

26

26

26

26

26

Maths*

100

101

97

97

100

Vocabulary*

54

68

40

50

100

Raven *

97

97

100

100

100

Experimental measures

Control

4

5

4

5

5

Visually similar

2

3

2

2.5

5

Phonologically similar

2

2

2

2.5

3

Sign similar

3

5

2

4

5

VSE

2

2

2

2.5

0

PSE

2

3

2

2.5

2

SSE

1

0

2

1

0

Segmenting (63)

0

40

0

36

60

Blending (15)

4

8

1

8

14

Phoneme deletion (10)

0

8

0

7

9

Letter-sound (50)

10

45

6

40

50

Word reading*

60

85

25

79

100

Spelling*

50

83

25

73

100

Nonword reading (10)

0

6

0

6

10

*standard scores

( ) total score possible

VSE = visual similarity effect

PSE = phonological similarity effect

SSE = sign similarity effect