Using computerized games to teach face recognition skills ...

JCPP

Journal Name

B 2 2 5 8

Manuscript No.

Journal of Child Psychology and Psychiatry **:* (2010), pp **?**

Dispatch: 13.4.10 Author Received:

Journal: JCPP CE: Vinoth Kumar No. of pages: 10 PE: Bhagyalakshmi doi:10.1111/j.1469-7610.2010.02258.x

1

2 3

Using computerized games to teach face

4

recognition skills to children with autism

5

6

spectrum disorder: The Let's Face It! program

7

8

9

James W. Tanaka,1 Julie M. Wolf,2 Cheryl Klaiman,2 Kathleen Koenig,2 Jeffrey

10

Cockburn,1 Lauren Herlihy,2 Carla Brown,2 Sherin Stahl,2 Martha D. Kaiser,3

11

and Robert T. Schultz4

12

1Department of Psychology, University of Victoria, British Columbia, Canada; 2Child Study Center, Yale University

13

School of Medicine, USA; 3Department of Psychology, Rutgers University, USA; 4Department of Pediatrics, University

14

of Pennsylvania School of Medicine, USA

15

16

17

Background: An emerging body of evidence indicates that relative to typically developing children,

18

children with autism are selectively impaired in their ability to recognize facial identity. A critical

19

question is whether face recognition skills can be enhanced through a direct training interven-

20

tion. Methods: In a randomized clinical trial, children diagnosed with autism spectrum disorder were

21

pre-screened with a battery of subtests (the Let's Face It! Skills battery) examining face and object processing abilities. Participants who were significantly impaired in their face processing abilities were

22

assigned to either a treatment or a waitlist group. Children in the treatment group (N = 42) received 20

23

hours of face training with the Let's Face It! (LFI!) computer-based intervention. The LFI! program is

24

comprised of seven interactive computer games that target the specific face impairments associated

25

with autism, including the recognition of identity across image changes in expression, viewpoint and

26

features, analytic and holistic face processing strategies and attention to information in the eye region.

27

Time 1 and Time 2 performance for the treatment and waitlist groups was assessed with the Let's Face

28

It! Skills battery. Results: The main finding was that relative to the control group (N = 37), children in

29

the face training group demonstrated reliable improvements in their analytic recognition of mouth

30

features and holistic recognition of a face based on its eyes features. Conclusion: These results indi-

31

cate that a relatively short-term intervention program can produce measurable improvements in the

32

face recognition skills of children with autism. As a treatment for face processing deficits, the Let's Face It! program has advantages of being cost-free, adaptable to the specific learning needs of the individual

33

child and suitable for home and school applications. Keywords: Face recognition, autism, computer-

34

based intervention, training, perceptual expertise. Abbreviations: LFI!: Let's Face It!; ASD: autism

35

spectrum disorder; RCT: randomized clinical trial; PDD-NOS: pervasive developmental disorder, not

36

otherwise specified; ADI-R: Autism Diagnostic Interview ? Revised; ADOS-G: Diagnostic Observation

37

Schedule ? Generic; WASI: Wechsler Abbreviated Scale of Intelligence; WISC-III: Wechsler Intelligence

38

Scale for Children; WAIS-III: Wechsler Adult Intelligence Scale; DAS: Differential Abilities Scales.

39

40

41

42 While it has been said that nearly everyone is an ism perform equally as well or better than age- and

43 expert in face recognition (Carey, 1992), this claim IQ-matched typically developing children on per-

44 might not be true for children with autism. An ceptual tasks involving non-face stimuli (e.g.,

45 emerging literature suggests that many individuals houses) (Wallace et al., 2008; Wolf et al., 2008). The

46 with autism spectrum disorder (ASD) are less likely research suggests that a substantial proportion of

47 to attend to faces (Swettenham et al., 1998), are individuals with ASD present significant and selec-

48 impaired in face discrimination tasks (Behrmann tive problems in their face recognition abilities.

49 et al., 2006; Tantam, Monaghan, Nicholson, & Stir- Indeed, recent neuro-cognitive theories of autism

50 ling, 1989; Wallace, Coleman, & Bailey, 2008) and suggest that impaired face processing might lie at

51 have difficulty recognizing familiar faces (Blair, Frith, the root of the social dysfunction of the disorder

52 Smith, Abell, & Cipolotti, 2002; Boucher & Lewis, (Dawson et al., 2005; Schultz, 2005).

53 1992; Gepner, de Gelder, & de Schonen, 1996;

An open question is whether the face recognition

54 Hauck, Fein, Maltby, Waterhouse, & Feinstein, skills of individuals with autism can be enhanced

55 1998; Klin et al., 1999). Difficulties in face process- through direct training. In other domains, percep-

56 ing are not attributable to deficits in basic visual or tual expertise training has proved effective for

57 perceptual impairments because children with aut- enhancing the abilities of neurotypical individuals to

58

recognize natural (Tanaka, Curran, & Sheinberg,

59

2005) and artificial objects, `Greebles' (Gauthier &

60

Conflict of interest statement: No conflicts declared.

Tarr, 1997). The expert training protocol emphasizes

? 2010 The Authors Journal compilation ? 2010 Association for Child and Adolescent Mental Health. Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA

2 James W. Tanaka et al.

1 the quick and accurate recognition of objects at 1999; Wolf et al., 2008) and failure to perceive faces

2 specific, subordinate levels of abstraction. In face holistically (Gauthier, Klaiman, & Schultz, 2009;

3 processing, expert training has also been shown to Joseph & Tanaka, 2003; Teunisse & de Gelder, 4 improve recognition of other-race faces where five 2003)1. The LFI! games are organized into a theo-

5 days of intense practice at individuating other-race retical hierarchy of face processing domains that

6 faces successfully ameliorates other-race face rec- reinforce the child's ability to attend to faces

7 ognition (Tanaka & Pierce, 2009). Clinically, face (Domain I), recognize facial identity and expression

8 training has been shown to remediate severe face (Domain II) and interpret facial cues in a social

9 recognition deficits (i.e., developmental prosopagn- context (Domain III) (Tanaka, Lincoln, & Hegg,

10 osia) in which performance is so compromised that 2003). Each game is designed with engaging graph-

11 the individual has difficulty recognizing familiar ics, an original music track, and at least 24 levels of

12 friends and relatives. In one study, 14 months of game play that become progressively more chal-

13 intensive face training improved a patient with lenging and complex (see Figure 2). For children with

14 developmental prosopagnosia's ability to discrimi- ASD, computer-based instruction has the benefit of

15 nate configural differences in a face and to recognize providing a stable, consistent learning environment

16 famous people (DeGutis, Bentin, Robertson, & that can be customized to the instructional needs of

17 D'Esposito, 2007). Collectively, these studies dem- the user (Moore, McGrath, & Thorpe, 2000). In other

18 onstrate the efficacy of perceptual expertise training areas, computer training and multi-technology have

19 procedures for enhancing the object and face recog- been shown to be successful for teaching emotional

20 nition abilities of healthy adults and patients with skills to children with autism (Silver & Oakes, 2001;

21 face processing disorders.

Golan & Baron-Cohen, 2006; Golan et al., in press).

22

Despite the positive face training results obtained

In the LFI! intervention study, children were pre-

23 with neurotypical adults and patients, few studies screened with Let's Face It! Skills Battery, a series of

24 have examined the effects of training to promote face measures that have been shown to be sensitive to the

25 expertise in children with autism. One exception is a face processing deficits in autism (Wolf et al., 2008).

26 recent study by Faja and colleagues (Faja, Aylward, Children with ASD who were significantly impaired

27 Bernier, & Dawson, 2008) where five male young on these measures relative to their age-matched

28 adults with ASD (mean age = 19.0 years, Full Scale peers qualified for the intervention study (see Meth-

29 IQ = 99.0) received training to identify Caucasian ods). Following the recommended guidelines for

30 faces according to age, gender and identity. After a assessing treatment efficacy (Lord et al., 2005), eli-

31 three-week training period, post-treatment results gible participants were randomly assigned to either

32 revealed that the training group showed greater an active treatment group (N = 42) or a waitlist con-

33 sensitivity to configural information (i.e., distances trol group (N = 37) in a randomized clinical trial

34 between the eyes) compared to untrained control (RCT). Children in the active treatment group played

35 participants. The Faja et al. results provide `proof of the LFI! games for an average of 20 hours in their

36 concept' that identity recognition skills can be home over a two- to four-month period. The child's

37 improved through practice in face recognition. game play was self-paced and not directly supervised

38 However, this approach has several limitations as a by the parent or caregiver. Performance was logged

39 general intervention for treating face processing on a file and sent on a weekly basis to the case

40 deficits in ASD. First, this was a relatively small managers to monitor treatment compliance. After 20

41 intervention study and the individualized treatment hours of treatment, the LFI! Skills Battery was

42 is less practical for a large-scale intervention. Sec- re-administered to the treatment and waitlist

43 ond, the protocol in Faja et al.'s paper does not participants. We hypothesized that children who

44 include any object condition. So, it is unclear whe- received the 20 hours of training with the LFI! pro-

45 ther the improved configural processing that they gram would show larger gains in their face process-

46 find for faces is in fact specific to this class of stimuli. ing abilities relative to children in the waitlist group

47 Third, the Faja et al. study focused on adult training as measured by the Let's Face It! Skills Battery.

48 which relied heavily on repeated presentations of the

49 same learning trials and speeded reaction time

50 responses. It is unlikely that this method would be Method

51

appropriate for children with ASD's limited attention

This study was approved by the institutional review

52 53

span and less reliable reaction times (Rinehart, Bradshaw, Brereton, & Tonge, 2001).

boards at both the Yale University School of Medicine

54

The Let's Face It! (LFI!) program comprises seven

and the University of Victoria. All participants (or parents of minor participants) gave written informed

55 56

interactive computer games that address the specific face processing deficits in autism, including inat-

consent after study procedures were fully explained

57

tention to the eyes (Rutherford, Clements, & Sekuler,

to them.

58 2007; Wolf et al., 2008), impaired recognition of

59

identity (Blair et al., 2002; Boucher & Lewis, 1992;

1 The Let's Face It! program can be downloaded free-of-charge

60

Gepner et al., 1996; Hauck et al., 1998; Klin et al.,

from the website: .

? 2010 The Authors Journal compilation ? 2010 Association for Child and Adolescent Mental Health.

Face training 3

1 2

Participants

Wechsler Adult Intelligence Scale, 3rd edition (WAIS-III; Wechsler, 1997), or the Differential Abilities Scales

3

Participants of the present study included 79 children,

(DAS; Elliott, 1990). In cases in which a participant had

4

adolescents, and young adults with autism spectrum

an IQ test administered clinically within the last year,

5

disorders. Participants in the ASD group were recruited

an IQ measure was not re-administered, and scores

6 7 8

on the basis of previous diagnoses of autistic disorder, Asperger's disorder, or pervasive developmental disorder, not otherwise specified (PDD-NOS), through presentations at schools and parent organizations, and

from the previous administration were utilized for the purposes of the present study.

Eligible participants were randomly assigned to either an active treatment group or a waitlist control

9

through existing relationships with families of children

group. Randomization was stratified by mental age

10

on the autism spectrum. Participants were excluded

(above or below 8 years) and diagnosis (autistic dis-

11

(a) if they had vision worse than 20?100 in both eyes,

order vs. autism spectrum (i.e., Asperger's disorder or

12

(b) if, in the judgment of an experienced clinician, they

PDD-NOS). For purposes of stratification, mental age

13

were unable to comprehend the instructions of the

was calculated as (chronological age * Full Scale IQ /

14

experimental tasks, or (c) if they did not have face

100). The active treatment group consisted of 42 par-

15

processing deficits significant enough to warrant

ticipants (34 males and 8 females) with a mean age of

16

intervention (inclusion required impairments of 2 or

10.5 (SD = 3.8) and a mean full scale IQ of 93.6 (SD =

17

more standard deviations below age-matched typically

22.1). The active treatment group comprised 27 indi-

18 19

developing controls on at least 33% of variables, or 1 or more standard deviations below controls on at least 50% of variables).

viduals with autistic disorder, 6 with Asperger's disorder, and 9 with PDD-NOS. The waitlist control group was composed of 37 children (28 males and 9 females)

20

Autism spectrum diagnoses were confirmed based on

with a mean age of 11.4 (SD = 3.7) years and a mean

21

DSM-IV criteria through use of the Autism Diagnostic

full scale IQ of 95.9 (SD = 23.4). The waitlist control

22

Interview ? Revised (ADI-R; Rutter, Le Couteur, & Lord,

group comprised 17 individuals with autistic disorder,

23

2003) and the Autism Diagnostic Observation Schedule

6 with Asperger's disorder, and 14 with PDD-NOS. The

24

? Generic (ADOS-G; Lord, Rutter, DiLavore, & Risi,

active treatment and waitlist control groups did not

25

1999) by a clinician trained in their administration,

significantly differ with regard to age (t(77) = 1.12, n.s.)

26

with at least five years of experience working with

or IQ (t(77) = .45, n.s.). Mean ADOS and ADI algorithm

27

individuals with autism spectrum disorders. In some

totals for the two groups are presented in Table 1.

28

cases, data were missing (ADOS: 2 missing; ADI: 5

29 30 31

missing), or participants did not meet criteria for an autism spectrum disorder on one of these measures (ADOS: 11 did not meet; ADI: 8 did not meet; note that there is no overlap in these numbers; i.e., all partici-

The Let's Face It! Skills Battery

The primary outcome measures for the present study were the facial identity and object processing subtests

32

pants met criteria on at least one of the two diagnostic

of the Let's Face It! (LFI!) Skills Battery. The LFI! Skills

33

measures). In these instances, a final diagnostic deci-

Battery is a comprehensive, computer-based battery

34

sion was made by consensus among two or more clini-

that assesses perception of facial identity across a

35

cians with at least five years of experience in the field of

broad range of face processing tasks (for more details

36 autism spectrum disorders, independent of any about the individual subtests, see Wolf et al., 2008).

37

knowledge of how the child performed on the study's

The subtests evaluate the child's ability to: 1) match

38

outcome measures.

faces across changes in expression and masked fea-

39

IQ was obtained for all participants using either the tures, 2) discriminate featural and configural changes

40

Wechsler Abbreviated Scale of Intelligence (WASI;

in faces, 3) recognize face features presented in isola-

41

Wechsler, 1999), the Wechsler Intelligence Scale for

tion and in the whole face, and 4) identify faces in an

42

Children, 3rd edition (WISC-III; Wechsler, 1991), the

old/new recognition task. The battery also includes two

43

44

Table 1 Mean ADOS and ADI algorithm scores for the active treatment and waitlist control groups. Note that 4 participants are

45

excluded from the ADI averages, because they were administered an alternate version of the ADI (due to their recent prior partic-

46

ipation in another study that utilized that version)

47

ADOS Module 1

ADOS Module 2

ADOS Module 3

ADOS Module 4

ADI

48

49

Active treatment

N

50

Communication total

51

Socialization total

1 7.00 9.00

7 6.14 9.71

27 5.22 9.78

6 4.67 9.00

37 17.24 25.08

52

Stereotyped Behaviors total

N/A

N/A

N/A

N/A

4.70

53

Waitlist control

54

N

Communication total

55

Socialization total

56

Stereotyped Behaviors total

57

ASD cutoffs

0 N/A N/A N/A

4 6.25 12.00 N/A

23 4.74 8.57 N/A

9 4.56 8.89 N/A

33 15.70 22.73

5.03

58

Communication

59

Socialization

Stereotyped Behaviors

60

2 4 N/A

3 4 N/A

2 4 N/A

2

8

4

10

N/A

3

? 2010 The Authors Journal compilation ? 2010 Association for Child and Adolescent Mental Health.

4 James W. Tanaka et al.

1 control subtests with non-face stimuli that assess the ment, allocation, follow-up and analysis phases of the

2

child's ability to discriminate featural and configural

Let's Face It! intervention.

3

changes in houses and their short-term recognition of

4

cars.

Active treatment. The Let's Face It! intervention is

5

composed of 7 computer games targeting various face

6 Procedure 7

processing skills, as described in Appendix I. Screenshots from two of the games are presented in Figure 2.

8 9 10

Participants came to the Yale Child Study Center for an initial, 2-day (Time 1) visit. During that visit, they were administered the Let's Face It! Skills Battery in addition to other neuropsychological and behavioral measures

Players were able to select mode and level of play. Computer-animated graphics and high-score tables were included within each game as incentives to increase motivation to engage in the intervention.

11

(e.g., IQ testing, diagnostic measures, and other

Participants assigned to the active treatment group

12

experimental measures not reported in the present

were provided with the Let's Face It! computer game

13

paper).

intervention to take home with them, and were

14

At the end of their Time 1 visit, upon confirmation of

instructed to play the games for at least 100 minutes

15

study eligibility (as described above), participants were

per week. Each week, parents sent the researchers log

16

randomly assigned to either the active treatment or

files (automatically generated by the software) that

17

waitlist control group. Participants assigned to the ac-

documented details about the participant's game play.

18

tive treatment group underwent the Let's Face It!

Based on these log files, researchers were able to

19 20

intervention (as described below) for an average of 20.2 (SD = 10.3) hours of intervention over an average period of 19.1 (SD = 7.3) weeks. Participants in the waitlist

monitor participants' compliance with the intervention. An assigned case manager provided parents with feedback about their child's game play and suggested

21

control group underwent treatment as usual for a

games for the participant.

22

comparable period of time. Following the intervention or

All participants received monetary compensation for

23

waitlist period, participants returned to the Yale Child

their game play. In addition, all families were provided

24

Study Center for a follow-up (Time 2) visit, at which time

with a set of plastic token reinforcers to use to increase

25 the LFI! Skills Battery was repeated to assess for motivation to comply with the intervention. The way in

26

intervention outcome. See Figure 1 for a diagram

which these tokens were used was individualized for

27

describing the progress of subjects through the enroll-

each participant and was implemented by parents in

28

29

LOW RESOLUTION FIG

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

Figure 1 Diagram showing subjects' progress through the enrollment, allocation, follow-up and analysis phases of 1

60 the Let's Face It! intervention study

? 2010 The Authors Journal compilation ? 2010 Association for Child and Adolescent Mental Health.

Face training 5

Colour online, B&W in print

1

(a)

(b)

2

3

4

5

6

7

8

9

10

11

12

13 Figure 2 Sample screen shots from two of the Let's Face It! games. (a) In the game Splash, faces appear at random

14

locations in the display and the child's task is to `splash' the faces matching the identity of the target face (e.g.,

15

Wendy). (b) In ZapIt, the child uses a face `launcher' to connect three faces of the same identity. As the difficulty level

16

increases, faces can vary across multiple cues, such as viewing angle, expression and clothing

17

18 consultation with their case manager. Participants had domization was stratified by mental age (above or

19 the option of participating in a `high-scores website' in below 8 years) and diagnosis (autistic disorder vs.

20 which their high scores for the game were posted on a autism spectrum (i.e., Asperger's disorder or PDD-

21

website (under a pseudonym to ensure confidentiality)

NOS). For purposes of stratification, mental age was

22

so that participants could compete against one another,

calculated as (chronological age * Full Scale IQ /

23

thus adding another incentive to increase and improve game play.

100). The active treatment group consisted of 42

24

Participants continued in the intervention until their

participants (34 males and 8 females) with a

25

total intervention time reached 20 hours. Note that a

mean age of 10.5 and a mean full scale IQ of 93.6.

26

small number of participants completed fewer than 20

The active treatment group comprised 27 individuals

27 hours of intervention, which is attributable to two with autistic disorder, 6 with Asperger's disorder,

28 causes: 1) Due to a technical error in the software's and 9 with PDD-NOS. The waitlist control group was

29 logging capability, some early study participants' logged composed of 37 children (28 males and 9 females)

30 time on game was inflated, so that at the time of their with a mean age of 11.4 years and a mean full scale

31 Time 2 visit (when the software logs indicated 20+ hours IQ of 95.9. The waitlist control group comprised 17

32

of game play), they had in fact played somewhat less

individuals with autistic disorder, 6 with Asperger's

33

than 20 hours (n = 3). Once this error was identified,

disorder, and 14 with PDD-NOS. The Time 1 to Time

34

total time on game for these participants was corrected,

2 analyses were Bonferroni adjusted for the multiple

35

and these lower total game play times are reflected in the overall average time on game reported here. The

comparisons to p < .05 (using a corrected alpha of

36

technical glitch in the software was corrected so that it

.007).

37

did not affect subsequent participants. 2) Some partic-

One of the LFI! Skills Battery subtests, Parts/

38 ipants discontinued game play midway through the Whole Identity, demonstrated a significant interac-

39 intervention period, but agreed to come back for a tion between group and timepoint (see Table 2).

40 follow-up visit so as not to lose their data entirely.

This subtest assesses the extent to which individ-

41

uals use a featural or holistic face recognition

42

strategy (Tanaka & Farah, 1993). In the Parts/

43 Results

44

Wholes task (see Figure 2), participants are presented with a whole study face and then, in a

45 Separate mixed factorial ANOVAs were conducted for two-alternative forced choice paradigm, are asked

46 each of the LFI! Skills Battery's facial identity sub- to identify a face part presented either in isolation

47 tests. In each of these analyses, the independent or in the whole face. On this subscale, recognition

48 variables were group (active treatment, waitlist con- of an isolated eye or mouth part from a study face

49 trol) and timepoint (Time 1, Time 2), and the provides an index of analytic processing whereas

50 dependent variable was total percentage accuracy on recognition of the part when tested in the whole face

51 the given subtest. To assess for intervention out- stimulus is a measure of holistic processing.

52 come, the comparison of interest was the interaction Results of the significant group ? timepoint inter-

53 between group and timepoint, in order to determine action for the Parts/Whole Identity task, F(1, 71) =

54 whether the active treatment group demonstrated 9.15, p < .003, are depicted in Figure 3. For the

55 significantly greater improvement from Time 1 to active treatment group, direct comparisons between

56 Time 2 than did the waitlist control group. Interac- Time 1 and Time 2 showed reliable gains in the part

57 tion effects for each of the LFI! Skills Battery identity mouth condition, F(1, 38) = 5.35, p < .05, and whole

58 subtests are depicted in Table 2. Eligible partici- eyes condition, F(1, 38) = 7.69, p < .001 (Figure 4).

59 pants were randomly assigned to either an active There was a trend toward reliable improvements

60 treatment group or a waitlist control group. Ran- after training for recognition of the part eyes,

? 2010 The Authors Journal compilation ? 2010 Association for Child and Adolescent Mental Health.

6 James W. Tanaka et al.

1

Table 2 Group ? Timepoint interaction effects for each of the Let's Face It! Skills Battery subtest

2

Time 1

Time 2

F

p

3

4

Face subtests

5

Face dimensions

Active: 72.1%

Active: 79.8%

F (1, 75) = .20

n.s.

6

Waitlist: 74.2%

Waitlist: 81.0%

Immediate memory for faces

Active: 43.1%

Active: 43.0%

F (1, 71) = .47

n.s

7

Waitlist: 46.3%

Waitlist: 49.2%

8

Matching identity

9

Masked features

10

11

Expression

12

Parts/Whole identity

13

Active: 52.1% Waitlist: 54.6% Active: 49.1% Waitlist: 56.4% Active: 58.3% Waitlist: 63.3%

Active: 58.1% Waitlist: 58.5% Active: 52.0% Waitlist: 59.2% Active: 64.1% Waitlist: 63.1%

F (1, 73) = .77 F (1, 75) = .00 F (1, 71) = 9.15

n.s. n.s. p = .003

14

Object subtests

15

House dimensions

Active: 67.1%

Active: 72.8%

F (1, 68) = .53

n.s.

16

Waitlist: 68.8%

Waitlist: 73.0%

Immediate memory for cars

Active: 49.4%

Active: 54.0%

F (1, 43) = .33

n.s.

17

Waitlist: 51.8%

Waitlist: 53.6%

18

19

20 p = .06, and whole mouths, p = 07. Of note, the p < .01, as shown in Figure 5. None of the group by

21 waitlist control group showed marginally higher timepoint interactions for the other face or object

22 accuracy at baseline than the active treatment subscales were reliable, p < .05.

23 group (t(72) = 1.95, p = .06). The reason for this

To investigate the relationship between improve-

24 baseline difference is unknown and likely reflects ment on the Parts/Whole Identity subtest and par-

25 chance error, given that participants were randomly ticipant characteristics, correlations were conducted

26 assigned to groups. In response to this baseline between the Time 2 minus Time 1 difference score

27 difference, a follow-up analysis was conducted (for the Parts/Whole Identity subtest total score) and

28 using groups that were matched at baseline. These each of the following participant characteristics: age,

29 matched groups were created by eliminating from IQ, ADOS Social algorithm total, ADOS Communi-

30 the analysis the waitlist participants who had the cation algorithm total, ADOS Communication +

31 highest scores at baseline, resulting in an average Social algorithm total, ADI Social algorithm total,

32 accuracy for both groups of 58% at Time 1. Using ADI Communication algorithm total, ADI Stereo-

33 these matched groups, the group ? timepoint typed Behaviors algorithm total, and total time on

34 interaction remained significant, F(1, 63) = 7.33, intervention (in minutes). Improvement from Time 1

35

36

(a)

37

38

39

40

41

42

43

44

45

46

47

(b)

(c)

48

49

50

51

52

53

54

55

56

57

Figure 3 The Parts/Whole Test: (a) Study face, (b) `Part' test condition and (c) `Whole face' test condition. Note that in 2

58

`part' and `whole face' test conditions, the target and foil items only differ with respect to the critical eye features.

59

Analytic processing is measured by the correct recognition of the part in the isolated test condition. Holistic pro-

60 cessing is measured by improved recognition in `whole face' test condition relative to the isolated `part' test condition

? 2010 The Authors Journal compilation ? 2010 Association for Child and Adolescent Mental Health.

L O WColour online, B&W in print LOW RESOLUTION FIG

Face training 7

1

LOW RESOLUTION FIG

2

3

4

5

6

7

8

9

10

11

12

13

Figure 4 Significant (p = .003) group ? timepoint interaction for the Parts /Whole Identity task. The active treatment 3 group showed reliable gains in their ability to recognize mouths in isolation (*p < .05) and to process eyes holistically

14

(**p < .01)

15

16

17

LOW RESOLUTION FIG

18

However, the local bias is not immutable to train-

19

ing and practice. Children who participated in the

20

Let's Face It! face program improved in the holistic

21

recognition of the eyes. The improvement in holistic

22

eye recognition is notable because individuals with

23

autism are more likely to ignore distinguishing

24

information in the eye region in favor of the less

25

informative region of the mouth (Rutherford et al.,

26

2007; Wolf et al., 2008). Thus, as measured by the

27

Parts/Whole test, the Let's Face It! intervention was

28

successful in redirecting the participants' attention

29

to the eyes and integrating eye information in the

30

Figure 5 Group ? timepoint interaction for the Parts/ 4 whole face.

31 Whole Identity subtest when groups were matched at

The reliable improvements of the participants in

32 baseline, F(1,63) = 7.326, p < .01. The active treatment the treatment group cannot be attributed to baseline

33 group showed a significant increase in scores from Time differences. When participants in the treatment

34

1 to Time 2 (p < .001), while the waitlist group showed

group and waitlist group were matched according to

35

no change

the Time 1 performance, individuals in the treatment

36

to Time 2 did not correlate significantly with any of

group nevertheless showed appreciable gains com-

37

these participant characteristics.

pared to the control group. The results were specific

38

to faces, indicating that the improvement was not a

39

general training effect. To our knowledge, this is the

40 Discussion

41

first time that it has been shown in a large-scale, randomized clinical trial that face recognition abili-

42 In this study, it was found that 20 hours of face ties of children with autism can be improved through

43 training with the Let's Face It! program was suffi- an unsupervised, computer-based intervention.

44 cient to improve the analytic and holistic face pro-

In contrast to the perceptual gains demonstrated

45 cessing skills of children in the treatment group as on the Parts/Wholes measure, participants in the

46 assessed by the Parts/Wholes Identity Test. The treatment group did not improve in their ability to

47 post-treatment results revealed that LFI! training detect featural and configural face changes (Dimen-

48 enhanced the recognition of both the eye and mouth sions test), to identify faces across changes in

49 face features. Interestingly, the largest improve- expression and orientation (Matching Identity tests)

50 ments were found in analytic recognition when the or to recognize faces over a short retention (Imme-

51 face parts were tested in isolation. This finding is diate Memory for Faces test) relative to the waitlist

52 compatible with other results showing that individ- group. The absence of an interaction effect might

53 uals with autism are biased toward an analytic reflect practice effects from Time 1 testing to Time 2

54 approach to face processing (Gauthier et al., 2009; testing or overall, developmental improvements. The

55 Joseph & Tanaka, 2003). This result provides fur- absence of training-specific effects on these other

56 ther evidence that as a general perceptual strategy, measures suggest that further modification and

57 individuals with ASD focus more on the details of a improvements of the program are required, such as

58 stimulus rather than its global properties (Behr- providing children with explicit, rule-based strate-

59 mann et al., 2006; Mottron, Dawson, Soulieres, gies for aiding recognition (Faja et al., 2008). It is also

60 Hubert, & Burack, 2006).

possible that the treatment dosage of 20 hours was

? 2010 The Authors Journal compilation ? 2010 Association for Child and Adolescent Mental Health.

8 James W. Tanaka et al.

1 not sufficient to promote across-the-board improve-

This material is available as part of the online

2 ments in face processing.

article from:

3

Despite its limitations, the Let's Face It! program



4 shows promise as an effective intervention tool and 10.1111/j.1469-7610.2010.02258.x

5 treatment alternative. From a practical standpoint,

Please note: Blackwell Publishing are not respon-

6 the computer-based treatment is cost-free, can eas- sible for the content or functionality of any supple-

7 ily be implemented in a home, school or clinical mentary materials supplied by the authors. Any

8 environment and available on multiple operating queries (other than missing material) should be

9 systems (i.e., Mac OSX, Windows). The program does directed to the corresponding author for the article.

10 not require direct supervision and can be customized

11 to the learning skills of the individual child. Although

12 the extent to which gains on computer-based pro- Author notes

13 14

grams directly translate into improved socialization skills is unknown, it is believed that the Let's Face It!

Robert T. Schultz is now at the University of Penn-

15

program provides an important bridge for children

sylvania, Cheryl Klaiman at the Children's Health

16

with ASD to the face processing skills that are critical

Council, Mikle South at Brigham Young University and Martha D. Kaiser at Rutgers University.

17 in real-world face-to-face interactions. Indeed, the

18 program is not intended to be a stand-alone treat-

19 ment or a substitute for human interaction, but Acknowledgements

20 might best used in conjunction with human inter-

21 ventionists (Rogers, 2000) who can reinforce the This study was funded by grants from the NIH (Studies

22 principles introduced in the Let's Face It! software.

to Advance Autism Research and Treatment), the

23

In summary, our results demonstrate that like James S. McDonnell Foundation, the National

24 other forms of perceptual expertise, face processes Science Foundation (#SBE-0542013) and the Na-

25 are amendable to the effects of training and practice. tional Science and Engineering Research Councils of

26 In this study, 20 hours of Let's Face It! training was Canada. The authors wish to thank the following

27 sufficient to boost the face recognition skills of chil- individuals who were instrumental in software pro-

28 dren with ASD. Although this is a non-trivial gramming, data collection, and/or data entry: Mikle

29 investment of time, it pales in comparison to the South, James McPartland, Jennifer Hetzke, Diane

30 lifetime of experience that children have with faces. Goudreau, Dave Swanson, Zena Rittenhouse, Megan

31 Yet, this relatively modest amount of face training Myers, Andy Auerbach, Daniel Grupe, and Malia

32 was enough to bring about measurable gains in Wong. We also wish to thank the participants and

33 analytic and holistic face recognition skills of chil- their families who made this research possible.

34 dren with ASD. While encouraging, these results are

Online access to the Let's Face It! Skills Battery

35 only a first step toward developing a comprehensive can be obtained by contacting James Tanaka at

36 curriculum in face processing. Further research is jtanaka@uvic.ca.

37 warranted to test the long-term benefits of face

38 training and the degree to which training transfers to

39 everyday social skills.

40

Correspondence to

41

42 Supplementary material

James Tanaka, Department of Psychology, Univer-

43 44 45

The following supplementary material is available for this article:

sity of Victoria, Victoria, BC, V8W 3P5, Canada; Email: jtanaka@uvic.ca; or Robert T. Schultz, Center for Autism Research, The Children's Hospital of

46

Appendix 1. Description of the Let's Face It! Philadelphia, 8th Floor Suite 860, Philadelphia, PA

47

games (Word document)

19104, USA; Email: schultzrt@chop.edu

48

49

50

Key points

51

52

? Children with autism have significant deficits in face recognition.

53

? As a form of perceptual expertise, it is hypothesized that face recognition skills can be improved through

54

practice and training.

55

? In a randomized clinical trial, impaired face recognition was ameliorated through 20 hours of computer-

56

based treatment with the Let's Face It! program.

57

? The Let's Face It! program is a practical intervention in face processing for children with ASD.

58

? The Let's Face It! program can be downloaded free of charge from the website:

59

naka/letsfaceit.

60

? 2010 The Authors Journal compilation ? 2010 Association for Child and Adolescent Mental Health.

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download