- Email: [email protected]
Independent dimensions of hand preference: Reliability of the factor structure and the handedness inventory
independent Preference: Structure
Dimensions Reliability
of Hand
of the Factor
and the Handedness
Inventory
Jacqueline Liederman Boston University
Jane M. Healey &fountSinai Medical Center
Healey, Liederman, and Gesch wind (1986) suggested that handedness is not a unidimensionol rrait. They devised (I questionnaire with 55 handedness items and 6 familial hand preference items and administered it to 290 adults. Their analysis identified four independent factors underlying hand preference. In the current study, the reliability of all but one of the items on the original inventory was established by retesting 83 of the original subjects. An additional analysis revealed that approximately 2% of the population could be considered to have ‘ambiguous” handedness in the sense that they indicated unstable hand preference according to the criteria suggested by Soper and Satz (1984). The reliability of the factor structure was evaluated by factor analyzing data from a new sample of 487 subjects, demographically similar to those in the original sample. All four factors reappeared with minor shifts in item loadings. The original and second samples were then combined, and four factors were again found, even when the sample was reconstructed so that there were equal numbers of males andfemales. On the basis of these new analyses our initial conclusion was confirmed: hand preference must be characterized along more than one dimension.
There is considerable agreement that laterality is not a single entity and that the various measures of laterality are not closely related to each other. For example, asymmetries of hand, eye, ear, and foot preference are found to cluster separately by factor analytic procedures (e.g., Dean, 1982; Porac &
Requests for reprints should be addressed to Jacqueline Liederman, Ph.D., Psychology DeMA 02215. Copies of the partment, Boston University, 64 Cummington Street, Boston, handedness questionnaire are available upon request from either author. 371
372
J. Liederman and J. M. Healy
Coren, 1981). From this it is often concluded that the determinants of hand, ear, eye and foot preference must be different. In contrast, it is traditionally believed that hand preference, per se, is a single entity, which can be characterized along a single dimension (e.g., Bryden, 1977; Hardyck & Petrinovich, 1977; Williams, 1986) and inherited as a single factor (Annett, 1985). Only recently, has it begun to be appreciated that handedness itself, also may be multidimensional, with different dimensions reflecting different kinds of hand movements and the neural systems which control those movements. Dean (1982) used a 49 item inventory, of which 34 items related to hand preference and 15 items related to ear, eye or foot preferences. Six factors emerged, two of which related to distinct aspects of handedness. Beukelaar and Kroonenberg (1983) used an inventory with only hand preference items and identified several independent factors. An additional two sets of data could be interpreted as demonstrating independent dimensions of hand preference, but the authors did not emphasize these interpretations (Annett, 1970; and Plato, Fox and Garruto, 1984). Healey, Liederman, and Geschwind (1986) used an extensive inventory of items restricted to hand preference and attempted to characterize the independent dimensions of handedness. They computed a principal components factor analysis and identified four independent factors underlying hand preference. Two of these factors represented activities for which both righthanded and non-right-handed individuals had relatively weak hand preferences; the other two factors represented activities for which subjects had strong hand preferences. Healey et al. (1986) argued that the finding that handedness was not a continuous variable along a single dimension had several important implications. First, theories of the inheritance of hand preference would have to accommodate to the notion that different dimensions of hand preference may be differently inherited. Second, Healey et al. (1986) suggested that their findings established the need for a revamping of the notion of “mixed” hand preference. They cautioned that mixed hand preference could be the result of two very different patterns of neural organization. Mixed hand preference between independent dimensions of hand preference, even within right-handers, may be the norm. In contrast, mixed patterns of hand preference within any particular factor may signal atypical neural organization. Thus different patterns of mixed hand preference may have different implications neuropsychologically. Their final point was that the separate neural systems which may govern the different dimensions of hand preference may mature at different times (cf. Liederman, 1983) and may account for the finding that young children show a more “mixed” pattern of manual preference than adults (Porac & Coren, 1981). Given the broad theoretical implications of the notion that handedness is not a unidimensional trait, the current study was devoted to an assessment of the stability of the original Healey et al. (1986) findings. The focus of this
Independent
Dimensions of Hand Preference
373
paper was to assess the reliability of the factor matrices by collecting data from a new, demographically similar sample, and repeating the same factor analysis. After that, the original and replication samples were combined so as to establish a final set of factors and factor loadings which can be used for future research. Finally, a separate analysis was done with a sample carefully controlled for gender, so that it could be shown that the four-factor solution was not specific to the skewed sample which was used in the original Healey et al. (1986) study. The second focus of this paper was to establish the test-retest reliability of the handedness inventory devised by Healey et al. (1986). Included in that analysis was an evaluation of the reliability of subjects’ reports of their family pattern of hand preference. Aside from establishing the reliability of the inventory, the test-retest data have certain theoretical implications. It has recently been suggested that there may be a special subgroup of individuals with mixed hand preference who have what has been termed “ambiguous” hand preference (Soper & Satz, 1984). These individuals are inconsistent about their hand preference for a particular task from one time to the next. They differ from individuals with “ambidextrous” hand preference, in that the latter are inconsistent about their hand preference across manual tasks, but for any particular hand preference task their preference is stable. Soper and Satz (1984) have suggested that ambiguous hand preference may be associated with extensive bilateral cerebral injury. Soper and Satz (1984) and Soper, Satz, Orsini, Henry, Zvi, and Schulman (in press) have suggested that neither hemisphere of the ambiguously handed individual is sufficiently intact for the expression of manual dominance. It has even been suggested that there may be a special subgroup of autistic patients who have ambiguous hand preference. The problem with this concept is that it is not known to what extent nonbrain damaged adult subjects are consistent with reference to their hand preference. Soper and Satz (1984) cite Palmer (1964) as support for the notion that the base rate of ambiguous handedness is low, perhaps approaching 0%. Therefore the stability of hand preference in a sample of college students within whom the incidence of extensive bilateral brain damage would be expected to be extremely close to zero was assessed.
METHOD
Materials
The handedness inventory devised by Healey et al. (1986) was employed. It consists of 55 hand preference items and 6 items which assess family handedness. The precise wording of the hand preference items is available in
374
J. Liederrnan and J. M. Healy
the previous report. (Copies of the inventory are available upon request.) Subjects indicated their side and degree of hand preference according to the following five-point scale: (1) “right always,” (2) “right preferred but sometimes use left,” (3) “no preference (can use either),” (4) “left preferred but sometimes use right,” (5) “left always.” The family preference items assessed the hand preference of the mother, father, and offspring, as well as brothers, sisters, and “other relatives.” Response categories were “right ,” “left, ” “ambidextrous,” or “not sure.” Subjects were encouraged to fill out the questionnaires at home rather than at school, so that they would be able to check with their relatives before filling out family items.
Sabjecrs Of the original 290 subjects who had previously filled out the questionnaire, requests for participation were mailed to all of the non-right-handers (N=80), and a random subgroup of right-handers (N=80), for whom we had address information. Forty questionnaires did not reach their destination due to change of address. Eighty-three subjects completed the inventory for a second time [56 self-declared right-handers (25 male, 31 female), 20 self-declared left-handers (7 male, 13 female), and 7 ambidexters (4 male, 3 female)].
RESULTS AND DISCUSSION Part I: Reliability of the Responses to the Handedness
Inventory
Hand preference items. McNemar (1947) tests for significance of changes were computed (cited in Siegel, 1956, pp. 63-67). The data were reduced so that a “1” or “2” response was coded as “right” and a “3”, “4” or “5” response was coded as “non-right.” The results for right-handers and non-right handers are summarized in Column 1 of Table 1. Results indicated that for all items except “to point to something in the distance,” subjects consistently indicated that they had a right or a non-right (i.e., a left or ambilateral) hand preference. This establishes that the responses to the handedness inventory were reliable except for that one item. The percentage of subjects that chose the same side response from Time 1 to Time 2 varied from 100% (for the hand preferred for writing) to 69% (for the hand preferred for pointing to something in the distance). The responses of non-right-handers were found to be significantly less
375
Independent Dimensions of Hand Preference TABLE Reliability 554em
of Responses on the Heal!,
Handedness
Questionnaire
for Self-declared
Non-right-handed
Hand Preference
Right-Handers
Write Blackboard Draw Throw ball Tennis Strike match Scissors Knife Hammer Screwdriver Brush teeth Thread needle Shave Fork Sew Corkscrew Peel apple Dart Gun Saw Letter opener Tweezers Wash dish Pour coffee Set clock Comb Paint wall Baseball bat Note: Numbers to Time 2.
Non-right Handers
100.0 100.0 100.0 98.2 98.2 98.2 100.0 96.4 100.0 98.2 96.4 94.6 96.4 96.4 96.4 92.3 98.2 96.4 100.0 98.2 94.6 96.4 98.2 98.2 89.3 92.1 98.1 83.3 indicate
the percentage
100.0 100.0 100.0 96.3 92.6 92.3 100.0 73.1 88.0 84.0 100.0 92.3 88.8 92.6 100.0
92.3 92.6 92.6 79.2 84.6 92.6 92.6 88.8 85.2 74.0 88.8 85.2 88.0
1 Liederman.
& Geschwind Right-handed
(1986) (,V = 56) and
subjects (R‘= 27)
Hand Preference
Right-Handers
Bowl Doorknob Shovel Type Calculator Color or paint Pour water Beat time Point Flip coin Snap Turn screw Wind watch Dial phone Catch Deal cards Bottle open Button phone Cartwheel Iron Axe Pound Lightbulb Erase Wave Erase blackboard Suitcase
of subjects
providing
consistent
98.2
82.1 85.5 100.0 89.3 100.0 89.3 12.7 67.9 98.2 67.9 83.6 91.1 91.1 87.5 96.4 91.1 84.0 88.8 96.3 88.7 84.0 91.1 94.6
76.8 72.1 71.4
responses
Non-right Handers 88.8 63.6 96.2 81.5 74.0 92.6 85.2 81.5 73.1 80.8 70.4 81.5 17.7 80.0 84.6 92.6 95.8 85.2 76.2 85.2 84.6 77.7 80.8 92.6 17.7 88.8 92.3
from Time 1
reliable than the responses of right-handers, across items: t(1, 108) = 3.12, subjects (N=7) were eliminated from the computation, the reliability of the self-declared left-handers and right-handers did not differ.
p< .Ol, however when the ambidextrous
Reliability of the Family Hand Preference Data. The family hand preference
data were reduced in several ways. First, ambidextrous and left-handed relatives were regrouped as “non-right-handed” relatives. This coding was chosen so as to be consistent with a popular definition of familial sinistrality
376
J. Liederman and J. M. Healy
(e.g., Boklage, 1981; Briggs, Nebes, & Kinsbourne, 1976; Orsini, Satz, Soper, & Light, 1985; Spiegler & Yeni-Komshian, 1983). Non-right-handed relatives were also treated as a single group because left-handers have relatively mixed hand preferences from task to task, and each informant would be apt to use a different criterion for distinguishing between left-handed and ambidextrous relatives. Second, there was a need to redefine the presence of familial sinistrality in the extended family because of the variable way that subjects responded to the question “Is there anyone else in your family (grandparents, aunts, etc.) who is left-handed or ambidextrous? If so, who?” Many subjects answered the first question “yes,” but left the second question blank. This made it impossible to tabulate the exact number of sinistral relatives in the extended family. To be conservative, therefore, irrespective of how many non-righthanded relatives the subject actually had, if a subject answered this question with a “yes”, the subject was only given credit for one non-right-handed relative in the extended family. Several different measures of familial sinistrality were then generated: 1) FS(e1) (at least one non-right-handed relative in the entire immediate or extended family); 2) FS(e2) (at least one non-right-handed relative in the immediate family and one in the extended family); 3) FS(i1) [at least one non-right-handed relative in the immediate family (i.e., among the parents, siblings, or offspring of the target person)]; 4) FS(i2) (at least two non-righthanded relatives in the immediate family). McNemar tests for significance of changes (Siegel, 1956) were used to compare results on the family handedness data from Time 1 to Time 2. Two types of reliability were computed. The first analyses computed whether a subject would be re-categorized in terms of presence or absence of “familial sinistrality” from Time 1 to Time 2, according to each of the four definitions of familial sinistrality. Results indicated that there were no significant changes in familial sinistrality categorizations from Time 1 to Time 2. Reliability was greatest for the strictest definition [FS(iZ); 940701and lowest for the most lenient definition [FS(el); 81%]. The second analysis examined the reliability from Time 1 to Time 2 of subjects’ reports for maternal left handedness (present vs. absent) and paternal left handedness (present vs. absent). The data were reduced to two categories: “right preference” and “non-right preference” (i.e., left or ambidextrous preference). Data were excluded from the computation if the subjects responded “don’t know” at Time 1 and/or Time 2. Once again these categorizations did not significantly differ from Time 1 to Time 2. Reliability was essentially the same for reports of maternal and paternal hand preference; 97% of responses were reliable.
Independent Dimensions of Hand Preference Part II:
377
An Estimate of ‘2mbiguous” Hand Preference in Our College
Sample As was reviewed in the introduction, Soper and Satz (1984) claimed that there is a separate category of hand preference which would be referred to as ambiguous hand preference. They used a battery which required autistic subjects to perform eight tasks. Each performance was requested three times within each of two sessions, in a random order. Any individual who switched hand preference for more than two items was classified as having ambiguous hand preference. To establish an adult baseline for comparison, the reliability of each of the 83 subjects for 13 items from the current battery was assessed. The items selected were the four top-loading items from the first three factors plus the only item on the fourth factor. The data from the combined factor analysis were used: (items l-5, 18, 29, 30, 34, 36, 37, 47, and 53). In the current study, 93% of the subjects (i.e., 77/83) indicated the same direction of hand preference from Time 1 to Time 2 for all but one item. Only two subjects switched preference for more than two items indicating an incidence of ambiguous hand preference of about 2% for this adult sample. The consistency of preferred direction of hand preference across sessions confirms that the top loading items on our four factors were highly reliable items. Part III: Replication Factor Analysis Subjects. The handedness inventory was administered to a new population of 712 subjects which was demographically similar to the initial sample in terms of education and age. The original sample consisted mainly of students enrolled at the City University of New York. The new sample consisted mainly of students enrolled at Boston University. Of these subjects, a subsample of 487 subjects was selected so that the factor analysis could be done with a sample similar to Healey et al.? (1986) in terms of gender and hand preference. This subsample was composed by means of the following steps: All of the responses from non-right-handers were included in the analysis, and then the first right-handers who participated were selected. This established a sample with the same approximate proportions of right-handers, left-handers and ambidexters as were in the Healey et al. (1986) sample. For this replication factor analysis, 28.7% of the sample was left-handed (N= 141), and 8% was ambidextrous (N= 40). For the original factor analysis, 28% of the sample was left-handed and 10% was ambidextrous. Although we did not directly control for gender, the two samples were virtually identical in terms of the proportion of the sample which was male: 39% for the original
378
J. Liederman and J. M. Healy
factor analysis, and 38% for the replication factor analysis. For the replication factor analysis, there were 186 males, 303 females, and five subjects who did not provide gender information. Results and discussion of the Replication Factor Analysis. By means of the PA1 Principle Components Analysis of SPSS, four factors were identified with eigenvalues greater than one. The eigenvalues for these factors were 36.8, 2.0, 1.3 and 1.1, respectively. The proportion of variance accounted for by each factor was 67%, 4%, 2% and 2070,respectively, so that together they accounted for 75% of the variance. A listing of the items with loadings greater than .60 from this replication factor analysis are presented in the second column of Table 2 along side the top loading items from the original Healey et al. (1986) study. By comparing columns 1 and 2, one can see that the replication factor analysisgroduced factor loadings which were extremely similar to those which appeared in the first study. Factors 3 and 4 switched places in terms of the proportions of the variance that they accounted for; this difference was less than 1%. For simplicity, the factors are listed in Table 2 in the order that they occurred in the final factor analysis (i.e., replication sample plus original sample-refer to the next section). This required that for the original data, factors 3 and 4 be presented in reversed order. Part IV Identification of the Items with the Strongest Factor Loadings to Represent the Four Independent Dimensions within Hand Preference (Original and Replications Samples Combined) Subjects. For this analysis, the original sample (N=290) and replication sample (N= 487) were combined for a total of 777 subjects. There were 488 self-declared right-handers (170 male, 313 female, 5 unidentified); 66 selfdeclared ambidexters (37 male, 29 female); and 223 left-handers (90 males, 133 females). Results and discussion. By means of the PA1 Principle Components Analysis of SPSS, once again, four factors were identified with eigenvalues of at least one. The eigenvalues for these factors were 37.7, 2.2, 1.2, and 1.0, respectively. The proportion of variance accounted for by each factor was 69%, 4%, 2% and 2%, respectively, so that together they accounted for 77% of the variance. As in the original study, two factors were very strongly lateralized (Factors 1 and 3 in this analysis, Factors 1 and 4 in the original analysis), whereas two factors were more weakly lateralized (Factors 2 and 4 in this analysis, Factors 2 and 3 in the original analysis). Factors 2 and 4 are weakly lateralized in somewhat different ways, especially for the non-righthanders. Factor 2 consists of items for which subjects report a relatively high
379
Independent Dimensions of Hand Preference TABLE Content
2
and Item Loadings Above .60 for the
Four Factors identified Combined
in the Original,
and Gender-Controlled
Original Sample
Replication Sample
Replication. Samples
Combined Sample
Gender Matched Sample
FACTOR I 1 Write 3 Draw 2 Write blackboard 34 Color picture 52 Erase with pencil I4 Fork* 22 Tweezers I1 Brush teeth 15 Sew 13 Shave 21 Open letter 12 Thread needle 26 Comb 48 Iron* 23 Wash dish 24 Pour coffee 17 Peel apple 27 Paint wall 35 Pour water 20 Saw* 9 Hammer 45 Bottle opener IO Screwdriver I6 Corkscrew 6 Match 5 Tennis 31 Shovel 7 Cut with scissors 19 Gun 54 Erase blackboard 32 Type
FACTOR 2 46 Button phone* 42 51 37 39 33 30 53 36
Dial phone* Lightbulb* Point Snap Calculator* Doorknob Wave Beat to music
.88 .88 .86 .85 .84 .78 .77 .76 .74 .74 .73 .70 .70 .70 .69 .68 .67 .67 .65 .64 .63 .6 .62 .6l .61 + + + + + +
.88 .88 .87 .86 .83 .78 .74 .77 .8l .73 .69 .70 .69 .67 .68 .69 .78 .74 .63 .73 .75 .67 .72 .68 .67 .63 .63 .62 .61 .60 +
.69 .66 .65
.89 .89 .87 .87 .85 .8l .78 .79 .82 .75 .73 .70 .7l .7l .67 .71 .78 .75 .66 .75 .74 .68 .72 .69 .68
.6tl .6l + + +
.6: .67 .60 + .6l
.75 .74 .66 .64 .64 + + + +
.6’5 .60 .69 .62 .68 .67 .65 .69
.:6 .63 .71 .62 .69 .68 .69 .67
.70 .69 .66 .66 + + .64 .67 .67
.64 .63
.64 .66
.68 .70
.68 .68
.88 .88 .86 .86 .84 .78 .76 .78 .79 .74 .71 .71 .69 .69 .68 .69 .75 .71 .63 .70 .71
FACTOR 3 I8 Dart 29 Bowling ball
380
.I Liederman and J. M. Healy TABLE 2 Continued
Gender Matched Sample
Original Sample
Replication Sample
.62 + +
.67 .67 +
.70 .65 .65
.70 .67 .62
55 Suitcase
.65 64
+ +
+ +
+ +
49 Cartwheel Axe* 47
.63 .61
.;f3
.:0
.:9
4 Throw ball 28 Baseball bat 5 Tennis racquet FACTOR 4 28 Baseball bat
Combined Sample
*Indicates intrinsically biased items which use tools which are themselves asymmetric or are dictated by cultural norms to be used asymmetrically. + Indicates an item with a factor loading < .60.
rate of “either hand” responses. Factor 4 consists of items for which subjects report a relatively high rate of preference for the otherwise non-preferred hand. These factors are characterized in terms of the distribution of left, right, or either responses for the top loading items in Fig. 1. The items which loaded above .60 on each of these factors are indicated in column 3 of Table 2. As can be seen by comparing these factors across the three analyses, the factor loadings were reasonably reliable across studies demonstrating that there are differential dimensions of hand preference. Part V Replication of the Factor Structure with a Subsample of the Population Matched for Gender It has often been suggested that males have a greater tendency toward mixed preference than females (see Dean, 1981). In fact, Dean (1982) has shown that this sex difference in lateral preference occurs only for certain dimensions of hand preference. Therefore the question arose as to whether the four factor solution which was repeatedly replicated in our two samples as well as the combined sample was an artifact of the sex ratio of our sample. The sample consisted of many more females than males. Thus, the final section of this report examined the factor structure for a sub-group of the combined sample, with equal numbers of male and female right-handers, left-handers and ambidexters. Subjects. From the combined sample (see Part IV above) of 777 subjects, a subsample was selected so that a) there were equal numbers of male and
Independent
Dimensions
of Hand Preference
381
female right-handers, left-handers and ambidexters, and b) the proportion of the sample that was right-handed, left-handed and ambidextrous matched that in the original report (62%, 28%, and lo%, respectively). The final gender-controlled sample consisted of 594 subjects (298 right-handers, 168 left-handers, and 58 ambidexters). By means of the PA1 Principle Components Analysis of SPSS, once again, even within this gender-controlled sample, four factors were identified with eigenvalues of at least one. The eigenvalues for these factors were 36.1, 2.3, 1.4, and 1.0, respectively. The proportion of variance accounted for by each factor was 66%, 4%, 3% and 2%, respectively, so that together they accounted for 75% of the variance. As can be seen in column four of Table 2, the factor loadings for these factors are virtually identical to those in the previous three analyses.
Results.
FACTOR 1 m m
R
FACTOR 2
RIGHT-HANDERS NON-RIGHT-HINDERS
E E HANDLOF RESbONSE
R
too90-
a
R
FACTOR 3
R
E L L E HAND OF RESPONSE
RIGHT- HANDERS NON-RIGHT-HANDERS
I I I E L L E R HAND OF RESPONSE FACTOR 4
R
R
m
RIGHT-HANDERS
m
NON-RIGHT-HANDERS
E L L E R IiAN OF RE%P&SE
FIGURE I. The distribution of Right (RI, Either (E) and Left (L) hand responses for the activities represented by Factors 1 to 4 in right and non-right handers. These data are from the combined original and replication samples (N= 777 subjects; 488 right-handers and 289 nonright-handers).
382
L Liederman and J. M. Healy
It should be noted that separate analyses were also done within gender (N= 297 subjects, per analysis) and the same four-factor solution emerged for men and women as has previously been reported.
SUMMARY
AND GENERAL
DISCUSSION
The Reliability of Hand Preference by Item vs. Individual The handedness inventory originally reported in Healey et al. (1986) has been shown to be a useful and reliable instrument. Only one item, “to point to something in the distance,” did not show test-retest reliability. Since this was a top item in Factor II of the original study, as well as in the analyses reported below, this calls into question the interpretation originally given to Factor II. It has been suggested that orienting to the right and pointing with the right hand involve left-hemispheric motor programs which later become elaborated into programs for language (Kinsbourne & Lempert, 1979). One could argue that the fact that subjects were inconsistent with reference to the hand that they prefer for pointing argues against that theory. A more likely alternative is that pointing to something in the distance may be very different from pointing to a nearby object as a way of referring to it. It would be interesting to see if this second kind of pointing is more consistently rightbiased in right-handed individuals. The current data also have implications for the construction of handedness instruments. These reliability data will eventually be used to select items for an abbreviated handedness instrument. The final questionnaire will have to be validated by comparing self-declared preferences against the subject’s actual performance, for each task (e.g., Raczkowski, Kalat, and Nebes, 1974). The questionnaire is quite likely to be valid as compared to performance data, since a literature review indicates that concordances between behavior and self-reported hand preference are above 90% for most items (Porac & Coren, 1981). The Reliability of the Data from Right vs. Non-Right-Handers Self-declared non-right-handers provided less reliable data than righthanders. This is reminiscent of effects previously reported for both hand preference and performance assessments (Humphrey, 195 1; Satz, Achenbath, & Fennell, 1967) as well as assessments of perceptual asymmetries (Hines, Fennell, Bowers, & Satz, 1980). These data must be interpreted with caution. Non-right-handers comprise a heterogenous group which includes ambidexters as well as left-handers. When left-handers were treated separately, their reliability was as good as that of right-handers.
Independent Dimensions of Hand Preference
383
Finally, the reliability data have implications for Soper and Satz’s (1984) concept of ambiguous hand preference. They operationally defined ambiguous hand preference, in children, as a shift of hand preference across trials for at least two out of eight items. To establish an adult baseline for comparison, the reliability for 13 items from the current battery was assessed for each of 83 subjects. The items selected were the top-loading items from the four independent factors of hand preference identified in this study. Ninetythree percent of the subjects (i.e., 77/83) indicated the same direction of hand preference from Time 1 to Time 2 for all but one item. In fact, only two subjects switched preference for more than two items. This suggests that Soper and Satz (1984) are essentially correct in assuming that the incidence of ambiguous hand preference in a presumably neurologically-intact adult sample would be close to zero; in this sample the incidence was about 2%. The reader should be reminded, however, that Soper and Satz (1984) were evaluating consistency of performance over a series of six repetitions of eight tasks whereas the current study evaluated consistency of preference over only two repetitions for 13 tasks. Thus the ultimate estimate of ambiguous handedness by Soper and Satz’s (1984) criteria will require that adult subjects be tested for consistency of performance on Soper and Satz’s (1984) hand preference battery. The Reliability of the Family Handedness Assessment Many studies rely upon questionnaire date for the assessment of familial handedness, yet there are few studies which examine the reliability of these responses, or the reliability of various family handedness designations (for an exception refer to Dean, 1979). Porac and Coren (1979) compared adults’ assessment of their parents’ hand preference to the parents’ assessment of their own hand preference. The offspring systematically underestimated the incidence of left-handedness in their parents. Thus, even though the current data indicate that the familial sinistrality categorizations were highly reliable over a period between one and three years, these data probably underestimated the actual incidence of left- or mixed-handedness in the family. The Reliability of the Factor Matrix Indicating Independent Dimensions of Hand Preference For each set of factor analyses, including the final analysis where the sample was composed of equal numbers of males and females, four factors were identified. This confirms the original assertion of Healey et al. (1986) that handedness is not unidimensional. As in the original study, hand preference represented by two of the factors were highly lateralized, whereas hand preferences for the other two factors were relatively weakly lateralized.
384
.J Liederman and J M. Healy
However, there was some fluctuation between analyses in the exact items which loaded on various factors. This suggests that the final characterization of hand preference must await an even larger scale study, perhaps including separate analyses of the factors according to self-declared hand preference. In addition, since Factor 2 contained the pointing item, the data should be collected with this item reworded to assess which hand the subject would use to point out something to another person. This inventory differed from those which have preceded it in that a special effort was made to include items which involve movement of the proximal or axial musculature. Across factor analyses (i.e., in the analyses with the original sample,-the replication sample, and the combined samples), one factor always appeared which could be interpreted as representing lateral preference for movement of the axial musculature. In the initial factor analysis, several items loaded on this factor, whereas in the current factor analyses, only one item appeared: “to do a cartwheel, which hand lands on the floor first.” CONCLUSION
In summary, our original interpretation of two of the four factors withstood a replication test. Factor 1 can still be interpreted as representing general handedness for activities which require relatively fine motor control and continual modification of a motor program as a consequence of the prior movement’s effect on the environment. The two new items which loaded on this factor (shovel and scissors) both require some precision of motor control and modification of the behavior over successive movements. Similarly, Factor 4 activities still can be interpreted as relying upon strength and balance more than dexterity and may be involved with the axial movement system. In addition, although the other two factors changed in terms of specific item content, they remained stable in terms of the distribution of lateralized responses. One was always highly lateralized (Factor 3) whereas the other (Factor 2) showed little asymmetry of lateral preference, even for right-handed subjects (see Fig. 1). This replicates the initial finding that even within right-handed subjects, there are significant differences in the extent to which the right hand is preferred for activities. Since the questionnaire and the factor analyses upon which these inferences were based have been shown to be reasonably reliable, this indicates that handedness must be considered a multidimensional rather than a unidimensional trait. Thus these results are in fundamental agreement with those of Beukelaar and Kroonenberg (1983), Dean (1982), Plato, Fox, and Garruto (1984), and Healey et al. (1986) all of whom have shown that hand preference can be characterized along more than one dimension. Future research should look at both pathological and nonpatholo-
Independent
Dimensions
of Hand Preference
385
gical influences (cf. Liederman & Coryell, 1982) on the establishment of these independent dimensions of hand preference. Such studies would examine the relationships among these independent patterns of manual preference and familial patterns of hand preference (e.g., Liederman & Healey, 1987); and various forms of cognitive and/or neural dysfunction (e.g., Dean, Schwartz, 8~ Smith, 1981). Acknowledgement-We would like to thank the subjects who participated in this experiment, especially those who had the patience to fill out the questionnaire for a second time. We would also like to thank the faculty who assisted in the distribution of questionnaires.
REFERENCES Annett, M. (1970). A classification of hand preference by association analysis. &/fish Journal of Psychology, 61, 303-321. Annett, M. (1985). Left, right, hand and brain: The right shiff theory. Hillsdale, New Jersey: Lawrence Erlbaum Associates. Beukelaar, L. J., & Kroonenberg, P. M. (1983). Towards a conceptualization of hand preference. British Journal of Psychology, 14, 33-45. Boklage, C. (1981). On the distribution of nonrighthandedness among twins and their families. Acta Genetic Medical Gemellol, 30, 167-187. Brig@, G., Nebes, R., & Kinsbourne, M. (1976). Intellectual differences in relation to personal and family handedness. Quarter/y Journal of Experimental Psychology, 28, 591-601. Bryden, M. P. (1977). Measuring handedness with questionnaires. Neuropsychologia, 13, 617-624. Dean, R. S. (1979). Cerebral laterality and verbal performance discrepancies in intelligence. Journal of School Psychology, 17, 145-150. Dean, R. S. (1981). Cerebral dominance and childhood learning disorders: Theoretical perspectives. School Psychology Review, 10, 373-380. Dean, R. S. (1982). Assessing patterns of lateral preference. Clinical NeuropsychologJ,, 4, 124-128. Dean, R. S., Schwartz, N. H., & Smith, L. S. (1981). Lateral preference patterns as a discriminator of learning difficulties. Journal of Consulting and Clinical Psvchology, 49, 227-235. Hardyck, C., & Petrinovich, L. F. (1977). Left-handedness. Psychological Bu//erin, 84, 385-404. Healey, J. M., Liederman, J., & Geschwind, N. M. (1986). Handedness is not a unidimensional trait. Correx, 22, 33-53. Hines, D., Fennell, E., Bowers, D., & Satz, P. (1980). Left-handers show greater test-retest variability in auditory and visual asymmetry. Bra/n and Language, 10, 208-21 I. Humphrey, M. E. (1951). Consistency of hand usage. British Journal of Educatronal Psvchology. 21, 214-225. Kinsbourne, M., & Lempert, H. (1979). Does left brain lateralization of speech arise from right-biased orienting to salient percepts? Human Development, 22, 270-276. Liederman, J. (1983). Mechanisms underlying discontinuities in the development of handedness. In G. Young, S. J. Segalowitz, C. Corter, and S. Trehub (Eds.), Manualspecia//;ation and the devdoping brain (pp. 71-92). New York: Academic Press. Liederman. J., & Coryell. J. (1982). The origin of left hand preference: Pathological and nonpathological influences. Neuropsychologia, 20, 721-725. Liederman, J., & Healey, J. M. (1987, February). The effect of familial sinistrality upon several
386
J. Liederman and J. M. Healy independenf dimensions ofhondpreference.
Paper presented at the meeting of the International Neuropsychological Society, Washington D.C. Orsini, D., Satz, P., Soper, H., & Light, R. (1985). The role of familial sinistrality in cerebral organization. Neuropsychologia, 23, 223-232. Palmer, D. (1964). Development of a differentiated handedness. Psychological Bullerin, 62,
257-273. Plato, C., Fox, K., & Garruto, R. (1984). Measures of lateral functional dominance: Hand dominance. Human Biology, 56, 259-275. Porac, C., & Coren, S. (1981). Laferulpreferences and human behavior. Ne\v York: SpringerVerlag. Porac, C., & Coren, S. (1979). A test of the validity of offspring’s report of parental handedness. Perceptual and Motor Skills, 49, 227-23 1. Raczkowski, D., Kalat, J., & Nebes, R. (1974). Reliability and validity of some handedness questionnaire items. Neuropsychologiu, 12, 43-47. Satz, P., Achenbach, K., & Fennell, E. (1967). Correlations between assessed manual laterality and predicted speech laterality in a normal population. Neuropsychologia, 5, 295-310. Siegel, J. (1956). Nonparametric statistics for the social sciences. McGraw Hill: New York. Soper, H. V., & Satz, P. (1984). Pathological left-handedness and ambiguous handedness: A new explanatory model. Neuropsychologia, 22, 51 I-515. Soper, H. V., Satz, P., Orsini, D., Henry, R., Zvi, J., & Schulman, M. (in press). Handedness patterns in autism suggest subtypes. Journal of Autism and Developmental Disorders. Spiegler, B., & Yeni-Komshian, G. (1983). Incidence of left-handed writing in a college population with reference to family patterns of hand preference. Neuropsychologia, 21, 65 l-659. Williams, S. (1986). Factor analysis of Edinburgh Handedness Inventory. Cortex, 22, 325-326.
Dimensions Reliability
of Hand
of the Factor
and the Handedness
Inventory
Jacqueline Liederman Boston University
Jane M. Healey &fountSinai Medical Center
Healey, Liederman, and Gesch wind (1986) suggested that handedness is not a unidimensionol rrait. They devised (I questionnaire with 55 handedness items and 6 familial hand preference items and administered it to 290 adults. Their analysis identified four independent factors underlying hand preference. In the current study, the reliability of all but one of the items on the original inventory was established by retesting 83 of the original subjects. An additional analysis revealed that approximately 2% of the population could be considered to have ‘ambiguous” handedness in the sense that they indicated unstable hand preference according to the criteria suggested by Soper and Satz (1984). The reliability of the factor structure was evaluated by factor analyzing data from a new sample of 487 subjects, demographically similar to those in the original sample. All four factors reappeared with minor shifts in item loadings. The original and second samples were then combined, and four factors were again found, even when the sample was reconstructed so that there were equal numbers of males andfemales. On the basis of these new analyses our initial conclusion was confirmed: hand preference must be characterized along more than one dimension.
There is considerable agreement that laterality is not a single entity and that the various measures of laterality are not closely related to each other. For example, asymmetries of hand, eye, ear, and foot preference are found to cluster separately by factor analytic procedures (e.g., Dean, 1982; Porac &
Requests for reprints should be addressed to Jacqueline Liederman, Ph.D., Psychology DeMA 02215. Copies of the partment, Boston University, 64 Cummington Street, Boston, handedness questionnaire are available upon request from either author. 371
372
J. Liederman and J. M. Healy
Coren, 1981). From this it is often concluded that the determinants of hand, ear, eye and foot preference must be different. In contrast, it is traditionally believed that hand preference, per se, is a single entity, which can be characterized along a single dimension (e.g., Bryden, 1977; Hardyck & Petrinovich, 1977; Williams, 1986) and inherited as a single factor (Annett, 1985). Only recently, has it begun to be appreciated that handedness itself, also may be multidimensional, with different dimensions reflecting different kinds of hand movements and the neural systems which control those movements. Dean (1982) used a 49 item inventory, of which 34 items related to hand preference and 15 items related to ear, eye or foot preferences. Six factors emerged, two of which related to distinct aspects of handedness. Beukelaar and Kroonenberg (1983) used an inventory with only hand preference items and identified several independent factors. An additional two sets of data could be interpreted as demonstrating independent dimensions of hand preference, but the authors did not emphasize these interpretations (Annett, 1970; and Plato, Fox and Garruto, 1984). Healey, Liederman, and Geschwind (1986) used an extensive inventory of items restricted to hand preference and attempted to characterize the independent dimensions of handedness. They computed a principal components factor analysis and identified four independent factors underlying hand preference. Two of these factors represented activities for which both righthanded and non-right-handed individuals had relatively weak hand preferences; the other two factors represented activities for which subjects had strong hand preferences. Healey et al. (1986) argued that the finding that handedness was not a continuous variable along a single dimension had several important implications. First, theories of the inheritance of hand preference would have to accommodate to the notion that different dimensions of hand preference may be differently inherited. Second, Healey et al. (1986) suggested that their findings established the need for a revamping of the notion of “mixed” hand preference. They cautioned that mixed hand preference could be the result of two very different patterns of neural organization. Mixed hand preference between independent dimensions of hand preference, even within right-handers, may be the norm. In contrast, mixed patterns of hand preference within any particular factor may signal atypical neural organization. Thus different patterns of mixed hand preference may have different implications neuropsychologically. Their final point was that the separate neural systems which may govern the different dimensions of hand preference may mature at different times (cf. Liederman, 1983) and may account for the finding that young children show a more “mixed” pattern of manual preference than adults (Porac & Coren, 1981). Given the broad theoretical implications of the notion that handedness is not a unidimensional trait, the current study was devoted to an assessment of the stability of the original Healey et al. (1986) findings. The focus of this
Independent
Dimensions of Hand Preference
373
paper was to assess the reliability of the factor matrices by collecting data from a new, demographically similar sample, and repeating the same factor analysis. After that, the original and replication samples were combined so as to establish a final set of factors and factor loadings which can be used for future research. Finally, a separate analysis was done with a sample carefully controlled for gender, so that it could be shown that the four-factor solution was not specific to the skewed sample which was used in the original Healey et al. (1986) study. The second focus of this paper was to establish the test-retest reliability of the handedness inventory devised by Healey et al. (1986). Included in that analysis was an evaluation of the reliability of subjects’ reports of their family pattern of hand preference. Aside from establishing the reliability of the inventory, the test-retest data have certain theoretical implications. It has recently been suggested that there may be a special subgroup of individuals with mixed hand preference who have what has been termed “ambiguous” hand preference (Soper & Satz, 1984). These individuals are inconsistent about their hand preference for a particular task from one time to the next. They differ from individuals with “ambidextrous” hand preference, in that the latter are inconsistent about their hand preference across manual tasks, but for any particular hand preference task their preference is stable. Soper and Satz (1984) have suggested that ambiguous hand preference may be associated with extensive bilateral cerebral injury. Soper and Satz (1984) and Soper, Satz, Orsini, Henry, Zvi, and Schulman (in press) have suggested that neither hemisphere of the ambiguously handed individual is sufficiently intact for the expression of manual dominance. It has even been suggested that there may be a special subgroup of autistic patients who have ambiguous hand preference. The problem with this concept is that it is not known to what extent nonbrain damaged adult subjects are consistent with reference to their hand preference. Soper and Satz (1984) cite Palmer (1964) as support for the notion that the base rate of ambiguous handedness is low, perhaps approaching 0%. Therefore the stability of hand preference in a sample of college students within whom the incidence of extensive bilateral brain damage would be expected to be extremely close to zero was assessed.
METHOD
Materials
The handedness inventory devised by Healey et al. (1986) was employed. It consists of 55 hand preference items and 6 items which assess family handedness. The precise wording of the hand preference items is available in
374
J. Liederrnan and J. M. Healy
the previous report. (Copies of the inventory are available upon request.) Subjects indicated their side and degree of hand preference according to the following five-point scale: (1) “right always,” (2) “right preferred but sometimes use left,” (3) “no preference (can use either),” (4) “left preferred but sometimes use right,” (5) “left always.” The family preference items assessed the hand preference of the mother, father, and offspring, as well as brothers, sisters, and “other relatives.” Response categories were “right ,” “left, ” “ambidextrous,” or “not sure.” Subjects were encouraged to fill out the questionnaires at home rather than at school, so that they would be able to check with their relatives before filling out family items.
Sabjecrs Of the original 290 subjects who had previously filled out the questionnaire, requests for participation were mailed to all of the non-right-handers (N=80), and a random subgroup of right-handers (N=80), for whom we had address information. Forty questionnaires did not reach their destination due to change of address. Eighty-three subjects completed the inventory for a second time [56 self-declared right-handers (25 male, 31 female), 20 self-declared left-handers (7 male, 13 female), and 7 ambidexters (4 male, 3 female)].
RESULTS AND DISCUSSION Part I: Reliability of the Responses to the Handedness
Inventory
Hand preference items. McNemar (1947) tests for significance of changes were computed (cited in Siegel, 1956, pp. 63-67). The data were reduced so that a “1” or “2” response was coded as “right” and a “3”, “4” or “5” response was coded as “non-right.” The results for right-handers and non-right handers are summarized in Column 1 of Table 1. Results indicated that for all items except “to point to something in the distance,” subjects consistently indicated that they had a right or a non-right (i.e., a left or ambilateral) hand preference. This establishes that the responses to the handedness inventory were reliable except for that one item. The percentage of subjects that chose the same side response from Time 1 to Time 2 varied from 100% (for the hand preferred for writing) to 69% (for the hand preferred for pointing to something in the distance). The responses of non-right-handers were found to be significantly less
375
Independent Dimensions of Hand Preference TABLE Reliability 554em
of Responses on the Heal!,
Handedness
Questionnaire
for Self-declared
Non-right-handed
Hand Preference
Right-Handers
Write Blackboard Draw Throw ball Tennis Strike match Scissors Knife Hammer Screwdriver Brush teeth Thread needle Shave Fork Sew Corkscrew Peel apple Dart Gun Saw Letter opener Tweezers Wash dish Pour coffee Set clock Comb Paint wall Baseball bat Note: Numbers to Time 2.
Non-right Handers
100.0 100.0 100.0 98.2 98.2 98.2 100.0 96.4 100.0 98.2 96.4 94.6 96.4 96.4 96.4 92.3 98.2 96.4 100.0 98.2 94.6 96.4 98.2 98.2 89.3 92.1 98.1 83.3 indicate
the percentage
100.0 100.0 100.0 96.3 92.6 92.3 100.0 73.1 88.0 84.0 100.0 92.3 88.8 92.6 100.0
92.3 92.6 92.6 79.2 84.6 92.6 92.6 88.8 85.2 74.0 88.8 85.2 88.0
1 Liederman.
& Geschwind Right-handed
(1986) (,V = 56) and
subjects (R‘= 27)
Hand Preference
Right-Handers
Bowl Doorknob Shovel Type Calculator Color or paint Pour water Beat time Point Flip coin Snap Turn screw Wind watch Dial phone Catch Deal cards Bottle open Button phone Cartwheel Iron Axe Pound Lightbulb Erase Wave Erase blackboard Suitcase
of subjects
providing
consistent
98.2
82.1 85.5 100.0 89.3 100.0 89.3 12.7 67.9 98.2 67.9 83.6 91.1 91.1 87.5 96.4 91.1 84.0 88.8 96.3 88.7 84.0 91.1 94.6
76.8 72.1 71.4
responses
Non-right Handers 88.8 63.6 96.2 81.5 74.0 92.6 85.2 81.5 73.1 80.8 70.4 81.5 17.7 80.0 84.6 92.6 95.8 85.2 76.2 85.2 84.6 77.7 80.8 92.6 17.7 88.8 92.3
from Time 1
reliable than the responses of right-handers, across items: t(1, 108) = 3.12, subjects (N=7) were eliminated from the computation, the reliability of the self-declared left-handers and right-handers did not differ.
p< .Ol, however when the ambidextrous
Reliability of the Family Hand Preference Data. The family hand preference
data were reduced in several ways. First, ambidextrous and left-handed relatives were regrouped as “non-right-handed” relatives. This coding was chosen so as to be consistent with a popular definition of familial sinistrality
376
J. Liederman and J. M. Healy
(e.g., Boklage, 1981; Briggs, Nebes, & Kinsbourne, 1976; Orsini, Satz, Soper, & Light, 1985; Spiegler & Yeni-Komshian, 1983). Non-right-handed relatives were also treated as a single group because left-handers have relatively mixed hand preferences from task to task, and each informant would be apt to use a different criterion for distinguishing between left-handed and ambidextrous relatives. Second, there was a need to redefine the presence of familial sinistrality in the extended family because of the variable way that subjects responded to the question “Is there anyone else in your family (grandparents, aunts, etc.) who is left-handed or ambidextrous? If so, who?” Many subjects answered the first question “yes,” but left the second question blank. This made it impossible to tabulate the exact number of sinistral relatives in the extended family. To be conservative, therefore, irrespective of how many non-righthanded relatives the subject actually had, if a subject answered this question with a “yes”, the subject was only given credit for one non-right-handed relative in the extended family. Several different measures of familial sinistrality were then generated: 1) FS(e1) (at least one non-right-handed relative in the entire immediate or extended family); 2) FS(e2) (at least one non-right-handed relative in the immediate family and one in the extended family); 3) FS(i1) [at least one non-right-handed relative in the immediate family (i.e., among the parents, siblings, or offspring of the target person)]; 4) FS(i2) (at least two non-righthanded relatives in the immediate family). McNemar tests for significance of changes (Siegel, 1956) were used to compare results on the family handedness data from Time 1 to Time 2. Two types of reliability were computed. The first analyses computed whether a subject would be re-categorized in terms of presence or absence of “familial sinistrality” from Time 1 to Time 2, according to each of the four definitions of familial sinistrality. Results indicated that there were no significant changes in familial sinistrality categorizations from Time 1 to Time 2. Reliability was greatest for the strictest definition [FS(iZ); 940701and lowest for the most lenient definition [FS(el); 81%]. The second analysis examined the reliability from Time 1 to Time 2 of subjects’ reports for maternal left handedness (present vs. absent) and paternal left handedness (present vs. absent). The data were reduced to two categories: “right preference” and “non-right preference” (i.e., left or ambidextrous preference). Data were excluded from the computation if the subjects responded “don’t know” at Time 1 and/or Time 2. Once again these categorizations did not significantly differ from Time 1 to Time 2. Reliability was essentially the same for reports of maternal and paternal hand preference; 97% of responses were reliable.
Independent Dimensions of Hand Preference Part II:
377
An Estimate of ‘2mbiguous” Hand Preference in Our College
Sample As was reviewed in the introduction, Soper and Satz (1984) claimed that there is a separate category of hand preference which would be referred to as ambiguous hand preference. They used a battery which required autistic subjects to perform eight tasks. Each performance was requested three times within each of two sessions, in a random order. Any individual who switched hand preference for more than two items was classified as having ambiguous hand preference. To establish an adult baseline for comparison, the reliability of each of the 83 subjects for 13 items from the current battery was assessed. The items selected were the four top-loading items from the first three factors plus the only item on the fourth factor. The data from the combined factor analysis were used: (items l-5, 18, 29, 30, 34, 36, 37, 47, and 53). In the current study, 93% of the subjects (i.e., 77/83) indicated the same direction of hand preference from Time 1 to Time 2 for all but one item. Only two subjects switched preference for more than two items indicating an incidence of ambiguous hand preference of about 2% for this adult sample. The consistency of preferred direction of hand preference across sessions confirms that the top loading items on our four factors were highly reliable items. Part III: Replication Factor Analysis Subjects. The handedness inventory was administered to a new population of 712 subjects which was demographically similar to the initial sample in terms of education and age. The original sample consisted mainly of students enrolled at the City University of New York. The new sample consisted mainly of students enrolled at Boston University. Of these subjects, a subsample of 487 subjects was selected so that the factor analysis could be done with a sample similar to Healey et al.? (1986) in terms of gender and hand preference. This subsample was composed by means of the following steps: All of the responses from non-right-handers were included in the analysis, and then the first right-handers who participated were selected. This established a sample with the same approximate proportions of right-handers, left-handers and ambidexters as were in the Healey et al. (1986) sample. For this replication factor analysis, 28.7% of the sample was left-handed (N= 141), and 8% was ambidextrous (N= 40). For the original factor analysis, 28% of the sample was left-handed and 10% was ambidextrous. Although we did not directly control for gender, the two samples were virtually identical in terms of the proportion of the sample which was male: 39% for the original
378
J. Liederman and J. M. Healy
factor analysis, and 38% for the replication factor analysis. For the replication factor analysis, there were 186 males, 303 females, and five subjects who did not provide gender information. Results and discussion of the Replication Factor Analysis. By means of the PA1 Principle Components Analysis of SPSS, four factors were identified with eigenvalues greater than one. The eigenvalues for these factors were 36.8, 2.0, 1.3 and 1.1, respectively. The proportion of variance accounted for by each factor was 67%, 4%, 2% and 2070,respectively, so that together they accounted for 75% of the variance. A listing of the items with loadings greater than .60 from this replication factor analysis are presented in the second column of Table 2 along side the top loading items from the original Healey et al. (1986) study. By comparing columns 1 and 2, one can see that the replication factor analysisgroduced factor loadings which were extremely similar to those which appeared in the first study. Factors 3 and 4 switched places in terms of the proportions of the variance that they accounted for; this difference was less than 1%. For simplicity, the factors are listed in Table 2 in the order that they occurred in the final factor analysis (i.e., replication sample plus original sample-refer to the next section). This required that for the original data, factors 3 and 4 be presented in reversed order. Part IV Identification of the Items with the Strongest Factor Loadings to Represent the Four Independent Dimensions within Hand Preference (Original and Replications Samples Combined) Subjects. For this analysis, the original sample (N=290) and replication sample (N= 487) were combined for a total of 777 subjects. There were 488 self-declared right-handers (170 male, 313 female, 5 unidentified); 66 selfdeclared ambidexters (37 male, 29 female); and 223 left-handers (90 males, 133 females). Results and discussion. By means of the PA1 Principle Components Analysis of SPSS, once again, four factors were identified with eigenvalues of at least one. The eigenvalues for these factors were 37.7, 2.2, 1.2, and 1.0, respectively. The proportion of variance accounted for by each factor was 69%, 4%, 2% and 2%, respectively, so that together they accounted for 77% of the variance. As in the original study, two factors were very strongly lateralized (Factors 1 and 3 in this analysis, Factors 1 and 4 in the original analysis), whereas two factors were more weakly lateralized (Factors 2 and 4 in this analysis, Factors 2 and 3 in the original analysis). Factors 2 and 4 are weakly lateralized in somewhat different ways, especially for the non-righthanders. Factor 2 consists of items for which subjects report a relatively high
379
Independent Dimensions of Hand Preference TABLE Content
2
and Item Loadings Above .60 for the
Four Factors identified Combined
in the Original,
and Gender-Controlled
Original Sample
Replication Sample
Replication. Samples
Combined Sample
Gender Matched Sample
FACTOR I 1 Write 3 Draw 2 Write blackboard 34 Color picture 52 Erase with pencil I4 Fork* 22 Tweezers I1 Brush teeth 15 Sew 13 Shave 21 Open letter 12 Thread needle 26 Comb 48 Iron* 23 Wash dish 24 Pour coffee 17 Peel apple 27 Paint wall 35 Pour water 20 Saw* 9 Hammer 45 Bottle opener IO Screwdriver I6 Corkscrew 6 Match 5 Tennis 31 Shovel 7 Cut with scissors 19 Gun 54 Erase blackboard 32 Type
FACTOR 2 46 Button phone* 42 51 37 39 33 30 53 36
Dial phone* Lightbulb* Point Snap Calculator* Doorknob Wave Beat to music
.88 .88 .86 .85 .84 .78 .77 .76 .74 .74 .73 .70 .70 .70 .69 .68 .67 .67 .65 .64 .63 .6 .62 .6l .61 + + + + + +
.88 .88 .87 .86 .83 .78 .74 .77 .8l .73 .69 .70 .69 .67 .68 .69 .78 .74 .63 .73 .75 .67 .72 .68 .67 .63 .63 .62 .61 .60 +
.69 .66 .65
.89 .89 .87 .87 .85 .8l .78 .79 .82 .75 .73 .70 .7l .7l .67 .71 .78 .75 .66 .75 .74 .68 .72 .69 .68
.6tl .6l + + +
.6: .67 .60 + .6l
.75 .74 .66 .64 .64 + + + +
.6’5 .60 .69 .62 .68 .67 .65 .69
.:6 .63 .71 .62 .69 .68 .69 .67
.70 .69 .66 .66 + + .64 .67 .67
.64 .63
.64 .66
.68 .70
.68 .68
.88 .88 .86 .86 .84 .78 .76 .78 .79 .74 .71 .71 .69 .69 .68 .69 .75 .71 .63 .70 .71
FACTOR 3 I8 Dart 29 Bowling ball
380
.I Liederman and J. M. Healy TABLE 2 Continued
Gender Matched Sample
Original Sample
Replication Sample
.62 + +
.67 .67 +
.70 .65 .65
.70 .67 .62
55 Suitcase
.65 64
+ +
+ +
+ +
49 Cartwheel Axe* 47
.63 .61
.;f3
.:0
.:9
4 Throw ball 28 Baseball bat 5 Tennis racquet FACTOR 4 28 Baseball bat
Combined Sample
*Indicates intrinsically biased items which use tools which are themselves asymmetric or are dictated by cultural norms to be used asymmetrically. + Indicates an item with a factor loading < .60.
rate of “either hand” responses. Factor 4 consists of items for which subjects report a relatively high rate of preference for the otherwise non-preferred hand. These factors are characterized in terms of the distribution of left, right, or either responses for the top loading items in Fig. 1. The items which loaded above .60 on each of these factors are indicated in column 3 of Table 2. As can be seen by comparing these factors across the three analyses, the factor loadings were reasonably reliable across studies demonstrating that there are differential dimensions of hand preference. Part V Replication of the Factor Structure with a Subsample of the Population Matched for Gender It has often been suggested that males have a greater tendency toward mixed preference than females (see Dean, 1981). In fact, Dean (1982) has shown that this sex difference in lateral preference occurs only for certain dimensions of hand preference. Therefore the question arose as to whether the four factor solution which was repeatedly replicated in our two samples as well as the combined sample was an artifact of the sex ratio of our sample. The sample consisted of many more females than males. Thus, the final section of this report examined the factor structure for a sub-group of the combined sample, with equal numbers of male and female right-handers, left-handers and ambidexters. Subjects. From the combined sample (see Part IV above) of 777 subjects, a subsample was selected so that a) there were equal numbers of male and
Independent
Dimensions
of Hand Preference
381
female right-handers, left-handers and ambidexters, and b) the proportion of the sample that was right-handed, left-handed and ambidextrous matched that in the original report (62%, 28%, and lo%, respectively). The final gender-controlled sample consisted of 594 subjects (298 right-handers, 168 left-handers, and 58 ambidexters). By means of the PA1 Principle Components Analysis of SPSS, once again, even within this gender-controlled sample, four factors were identified with eigenvalues of at least one. The eigenvalues for these factors were 36.1, 2.3, 1.4, and 1.0, respectively. The proportion of variance accounted for by each factor was 66%, 4%, 3% and 2%, respectively, so that together they accounted for 75% of the variance. As can be seen in column four of Table 2, the factor loadings for these factors are virtually identical to those in the previous three analyses.
Results.
FACTOR 1 m m
R
FACTOR 2
RIGHT-HANDERS NON-RIGHT-HINDERS
E E HANDLOF RESbONSE
R
too90-
a
R
FACTOR 3
R
E L L E HAND OF RESPONSE
RIGHT- HANDERS NON-RIGHT-HANDERS
I I I E L L E R HAND OF RESPONSE FACTOR 4
R
R
m
RIGHT-HANDERS
m
NON-RIGHT-HANDERS
E L L E R IiAN OF RE%P&SE
FIGURE I. The distribution of Right (RI, Either (E) and Left (L) hand responses for the activities represented by Factors 1 to 4 in right and non-right handers. These data are from the combined original and replication samples (N= 777 subjects; 488 right-handers and 289 nonright-handers).
382
L Liederman and J. M. Healy
It should be noted that separate analyses were also done within gender (N= 297 subjects, per analysis) and the same four-factor solution emerged for men and women as has previously been reported.
SUMMARY
AND GENERAL
DISCUSSION
The Reliability of Hand Preference by Item vs. Individual The handedness inventory originally reported in Healey et al. (1986) has been shown to be a useful and reliable instrument. Only one item, “to point to something in the distance,” did not show test-retest reliability. Since this was a top item in Factor II of the original study, as well as in the analyses reported below, this calls into question the interpretation originally given to Factor II. It has been suggested that orienting to the right and pointing with the right hand involve left-hemispheric motor programs which later become elaborated into programs for language (Kinsbourne & Lempert, 1979). One could argue that the fact that subjects were inconsistent with reference to the hand that they prefer for pointing argues against that theory. A more likely alternative is that pointing to something in the distance may be very different from pointing to a nearby object as a way of referring to it. It would be interesting to see if this second kind of pointing is more consistently rightbiased in right-handed individuals. The current data also have implications for the construction of handedness instruments. These reliability data will eventually be used to select items for an abbreviated handedness instrument. The final questionnaire will have to be validated by comparing self-declared preferences against the subject’s actual performance, for each task (e.g., Raczkowski, Kalat, and Nebes, 1974). The questionnaire is quite likely to be valid as compared to performance data, since a literature review indicates that concordances between behavior and self-reported hand preference are above 90% for most items (Porac & Coren, 1981). The Reliability of the Data from Right vs. Non-Right-Handers Self-declared non-right-handers provided less reliable data than righthanders. This is reminiscent of effects previously reported for both hand preference and performance assessments (Humphrey, 195 1; Satz, Achenbath, & Fennell, 1967) as well as assessments of perceptual asymmetries (Hines, Fennell, Bowers, & Satz, 1980). These data must be interpreted with caution. Non-right-handers comprise a heterogenous group which includes ambidexters as well as left-handers. When left-handers were treated separately, their reliability was as good as that of right-handers.
Independent Dimensions of Hand Preference
383
Finally, the reliability data have implications for Soper and Satz’s (1984) concept of ambiguous hand preference. They operationally defined ambiguous hand preference, in children, as a shift of hand preference across trials for at least two out of eight items. To establish an adult baseline for comparison, the reliability for 13 items from the current battery was assessed for each of 83 subjects. The items selected were the top-loading items from the four independent factors of hand preference identified in this study. Ninetythree percent of the subjects (i.e., 77/83) indicated the same direction of hand preference from Time 1 to Time 2 for all but one item. In fact, only two subjects switched preference for more than two items. This suggests that Soper and Satz (1984) are essentially correct in assuming that the incidence of ambiguous hand preference in a presumably neurologically-intact adult sample would be close to zero; in this sample the incidence was about 2%. The reader should be reminded, however, that Soper and Satz (1984) were evaluating consistency of performance over a series of six repetitions of eight tasks whereas the current study evaluated consistency of preference over only two repetitions for 13 tasks. Thus the ultimate estimate of ambiguous handedness by Soper and Satz’s (1984) criteria will require that adult subjects be tested for consistency of performance on Soper and Satz’s (1984) hand preference battery. The Reliability of the Family Handedness Assessment Many studies rely upon questionnaire date for the assessment of familial handedness, yet there are few studies which examine the reliability of these responses, or the reliability of various family handedness designations (for an exception refer to Dean, 1979). Porac and Coren (1979) compared adults’ assessment of their parents’ hand preference to the parents’ assessment of their own hand preference. The offspring systematically underestimated the incidence of left-handedness in their parents. Thus, even though the current data indicate that the familial sinistrality categorizations were highly reliable over a period between one and three years, these data probably underestimated the actual incidence of left- or mixed-handedness in the family. The Reliability of the Factor Matrix Indicating Independent Dimensions of Hand Preference For each set of factor analyses, including the final analysis where the sample was composed of equal numbers of males and females, four factors were identified. This confirms the original assertion of Healey et al. (1986) that handedness is not unidimensional. As in the original study, hand preference represented by two of the factors were highly lateralized, whereas hand preferences for the other two factors were relatively weakly lateralized.
384
.J Liederman and J M. Healy
However, there was some fluctuation between analyses in the exact items which loaded on various factors. This suggests that the final characterization of hand preference must await an even larger scale study, perhaps including separate analyses of the factors according to self-declared hand preference. In addition, since Factor 2 contained the pointing item, the data should be collected with this item reworded to assess which hand the subject would use to point out something to another person. This inventory differed from those which have preceded it in that a special effort was made to include items which involve movement of the proximal or axial musculature. Across factor analyses (i.e., in the analyses with the original sample,-the replication sample, and the combined samples), one factor always appeared which could be interpreted as representing lateral preference for movement of the axial musculature. In the initial factor analysis, several items loaded on this factor, whereas in the current factor analyses, only one item appeared: “to do a cartwheel, which hand lands on the floor first.” CONCLUSION
In summary, our original interpretation of two of the four factors withstood a replication test. Factor 1 can still be interpreted as representing general handedness for activities which require relatively fine motor control and continual modification of a motor program as a consequence of the prior movement’s effect on the environment. The two new items which loaded on this factor (shovel and scissors) both require some precision of motor control and modification of the behavior over successive movements. Similarly, Factor 4 activities still can be interpreted as relying upon strength and balance more than dexterity and may be involved with the axial movement system. In addition, although the other two factors changed in terms of specific item content, they remained stable in terms of the distribution of lateralized responses. One was always highly lateralized (Factor 3) whereas the other (Factor 2) showed little asymmetry of lateral preference, even for right-handed subjects (see Fig. 1). This replicates the initial finding that even within right-handed subjects, there are significant differences in the extent to which the right hand is preferred for activities. Since the questionnaire and the factor analyses upon which these inferences were based have been shown to be reasonably reliable, this indicates that handedness must be considered a multidimensional rather than a unidimensional trait. Thus these results are in fundamental agreement with those of Beukelaar and Kroonenberg (1983), Dean (1982), Plato, Fox, and Garruto (1984), and Healey et al. (1986) all of whom have shown that hand preference can be characterized along more than one dimension. Future research should look at both pathological and nonpatholo-
Independent
Dimensions
of Hand Preference
385
gical influences (cf. Liederman & Coryell, 1982) on the establishment of these independent dimensions of hand preference. Such studies would examine the relationships among these independent patterns of manual preference and familial patterns of hand preference (e.g., Liederman & Healey, 1987); and various forms of cognitive and/or neural dysfunction (e.g., Dean, Schwartz, 8~ Smith, 1981). Acknowledgement-We would like to thank the subjects who participated in this experiment, especially those who had the patience to fill out the questionnaire for a second time. We would also like to thank the faculty who assisted in the distribution of questionnaires.
REFERENCES Annett, M. (1970). A classification of hand preference by association analysis. &/fish Journal of Psychology, 61, 303-321. Annett, M. (1985). Left, right, hand and brain: The right shiff theory. Hillsdale, New Jersey: Lawrence Erlbaum Associates. Beukelaar, L. J., & Kroonenberg, P. M. (1983). Towards a conceptualization of hand preference. British Journal of Psychology, 14, 33-45. Boklage, C. (1981). On the distribution of nonrighthandedness among twins and their families. Acta Genetic Medical Gemellol, 30, 167-187. Brig@, G., Nebes, R., & Kinsbourne, M. (1976). Intellectual differences in relation to personal and family handedness. Quarter/y Journal of Experimental Psychology, 28, 591-601. Bryden, M. P. (1977). Measuring handedness with questionnaires. Neuropsychologia, 13, 617-624. Dean, R. S. (1979). Cerebral laterality and verbal performance discrepancies in intelligence. Journal of School Psychology, 17, 145-150. Dean, R. S. (1981). Cerebral dominance and childhood learning disorders: Theoretical perspectives. School Psychology Review, 10, 373-380. Dean, R. S. (1982). Assessing patterns of lateral preference. Clinical NeuropsychologJ,, 4, 124-128. Dean, R. S., Schwartz, N. H., & Smith, L. S. (1981). Lateral preference patterns as a discriminator of learning difficulties. Journal of Consulting and Clinical Psvchology, 49, 227-235. Hardyck, C., & Petrinovich, L. F. (1977). Left-handedness. Psychological Bu//erin, 84, 385-404. Healey, J. M., Liederman, J., & Geschwind, N. M. (1986). Handedness is not a unidimensional trait. Correx, 22, 33-53. Hines, D., Fennell, E., Bowers, D., & Satz, P. (1980). Left-handers show greater test-retest variability in auditory and visual asymmetry. Bra/n and Language, 10, 208-21 I. Humphrey, M. E. (1951). Consistency of hand usage. British Journal of Educatronal Psvchology. 21, 214-225. Kinsbourne, M., & Lempert, H. (1979). Does left brain lateralization of speech arise from right-biased orienting to salient percepts? Human Development, 22, 270-276. Liederman, J. (1983). Mechanisms underlying discontinuities in the development of handedness. In G. Young, S. J. Segalowitz, C. Corter, and S. Trehub (Eds.), Manualspecia//;ation and the devdoping brain (pp. 71-92). New York: Academic Press. Liederman. J., & Coryell. J. (1982). The origin of left hand preference: Pathological and nonpathological influences. Neuropsychologia, 20, 721-725. Liederman, J., & Healey, J. M. (1987, February). The effect of familial sinistrality upon several
386
J. Liederman and J. M. Healy independenf dimensions ofhondpreference.
Paper presented at the meeting of the International Neuropsychological Society, Washington D.C. Orsini, D., Satz, P., Soper, H., & Light, R. (1985). The role of familial sinistrality in cerebral organization. Neuropsychologia, 23, 223-232. Palmer, D. (1964). Development of a differentiated handedness. Psychological Bullerin, 62,
257-273. Plato, C., Fox, K., & Garruto, R. (1984). Measures of lateral functional dominance: Hand dominance. Human Biology, 56, 259-275. Porac, C., & Coren, S. (1981). Laferulpreferences and human behavior. Ne\v York: SpringerVerlag. Porac, C., & Coren, S. (1979). A test of the validity of offspring’s report of parental handedness. Perceptual and Motor Skills, 49, 227-23 1. Raczkowski, D., Kalat, J., & Nebes, R. (1974). Reliability and validity of some handedness questionnaire items. Neuropsychologiu, 12, 43-47. Satz, P., Achenbach, K., & Fennell, E. (1967). Correlations between assessed manual laterality and predicted speech laterality in a normal population. Neuropsychologia, 5, 295-310. Siegel, J. (1956). Nonparametric statistics for the social sciences. McGraw Hill: New York. Soper, H. V., & Satz, P. (1984). Pathological left-handedness and ambiguous handedness: A new explanatory model. Neuropsychologia, 22, 51 I-515. Soper, H. V., Satz, P., Orsini, D., Henry, R., Zvi, J., & Schulman, M. (in press). Handedness patterns in autism suggest subtypes. Journal of Autism and Developmental Disorders. Spiegler, B., & Yeni-Komshian, G. (1983). Incidence of left-handed writing in a college population with reference to family patterns of hand preference. Neuropsychologia, 21, 65 l-659. Williams, S. (1986). Factor analysis of Edinburgh Handedness Inventory. Cortex, 22, 325-326.