2

Cognition, 17 (1984) 137-153

Linking words and things: Basic processes and individual variation*

guage Acquisition. Cambridge, Mass., MIT

icquisitional data. In N. Hornstein and D.

Problem of Language Acquisition, 241-271,

chacun son du, nous presentons un modele e est reliee aux propositions sur la structure ntral de la theorie linguistique. Aussi Ie fait semble violee au cours de certaines experi­ qu'on observe dans les experiences n'est pas onsun langagede processus pour representer enes dans Ie developpement, en incluant les ;suggereespar notre approche. L'explication Itexecuter un plan. Comme cette explication ants n'a pas a etre tenue pour fautive. La syntagmatiques disponibles pour les enfants

WILMA BUCCI Adelphi University

Abstract This study investigates individual differences in activity or strength ofthe system of referential connections between verbal and nonverbal representations in the mind. Level of Referential Activity (RA) as measured by direct naming speed is related to performance in two more complex verbal performance tasks: (1) generating descriptive terms to distinguish closely related colors; and (2) generating brief spoken descriptions of a personal experience. Subjects with rapid naming speeds were more likely to produce spoken narratives high in qualities of immediacy and objectivity, and also to use a metaphoric style in generating color terms, i.e., by using terminology referring to entities outside of the color domain. Performance in both tasks indicated greater use of 'image to word' links in the high RA subjects; more reliance on 'word to word' links in the low RA group. The function involved in RA is distinct from the type of abstract ability measured by standard verbal intelligence and fluency tests, and is also distinct from imagery ability per se. In studies of the mental representations of words and things and the linkage between them, the significance of individual differences has often been ig­ nored. This has been true both in developmental studies and in experimental investigations with adult subjects. This paper focuses on individual differ­ ences in the referential process-the process of connecting nonverbal rep­ resentations to words. However we are primarily concerned not with the issue of individual variation per se, but with using such variation as a means of increasing our understanding of the referential process itself. *Research for this study was carried out in the Clinical Commucation Research Unit, Downstate Medical Center, S.U.N.Y. The author wishes to thank Dr. Norbert Freedman, Director of the Unit, for making the research possible, and Drs. Richard Billow, Felix Barroso, and Zulfiqar GHani for their valuable comments on an earlier version of this paper. Reprint requests should be sent to Wilma Bucci, Institute of Advanced Psychological Studies, Adelphi University, Garden City, NY 11530, U.S.A.

0010-0277/84/$5.60

© Elsevier Sequoia/Printed in The Netherlands

138

W. Bucci

The significance of individual differences in verbal and imagery abilities has been considered by Paivio and others in the context of the dual code approach (Paivio, 1971, 1982). This model incorporates two independent representational systems, a verbal system specialized for linguistic informa­ tion, and an imagery system representing perceptual information, with a com­ plex system of referential connections between the two. The referential links are most direct for concrete words and the entities which they represent; less direct for abstract concepts and terms. In a number of studies, Paivo and his colleagues have attempted to show that variation in verbal and imagery abilities affects performance in certain experimental tasks. For example, Ernest and Paivio (1969) found a relation­ ship between imagery ability, as measured by spatial manipulation tests, and incidental memory in a paired associates learning task. However the findings as to this effect are somewhat unclear and later studies showed little or no relationship between the same tests and other memory tasks (Paivio, 1982). The findings as to the functional significance of individual differences have been more consistent in studies involving symbolic comparison tasks. Subjects with high imagery ability had faster reaction times on a mental clocks compari­ son measure; this involves comparing sizes of the angles formed by the hands of two clocks, given the times in digital form (Paivio, 1978a). High imagers were also significantly faster than low imagers in comparing pictures and words on a pleasantness-unpleasantness dimension; the effect of verbal abil­ ity approached significance (Paivio, 1978b). (Paivio postulates that informa­ tion associated with affective dimensions is represented more directly in the nonverbal system). The findings in the latter study were also complicated by interaction be­ tween the two ability domains. The effect of imagery ability on performance was greater when verbal ability was high; reciprocally, verbal ability was effective only when imagery ability was high. In our terms, the data suggest the possibility that some of the performance variance attributed to either verbal or imagery ability may in fact be a function of variation in degree of integration or linking of the two symbolic domains. The possibility of indi­ vidual differences in activity of the system of referential links has not been directly investigated by Paivio; however it seems consistent with the premises of the dual code approach. This paper investigates individual differences in the activity of the referen­ tial system, i.e., variation between individuals in activity or strength of the referential links between verbal and nonverbal representations in the mind. This may be expected to affect verbal performance on a variety of different levels. The most direct manifestation of the referential linking function is simple naming itself--entering semantic memory with a particular entity to

Linking words and things

bal and imagery abilities context of the dual code oorates two independent ~d for linguistic informa­ l information, with a com­ two. The referential links which they represent; less

; have attempted to show ts performance in certain o (1969) found a relation­ al manipulation tests, and ask. However the findings tudies showed little or no nary tasks (Paivio, 1982). ndividual differences have omparison tasks. Subjects n a mental clocks compari­ ngles formed by the hands via, 1978a). High imagers 1 comparing pictures and 1; the effect of verbal abil­ a postulates that informa­ ented more directly in the

olicated by interaction be­ ery ability on performance .ocally, verbal ability was IUr terms, the data suggest ~iance attributed to either n of variation in degree of ns. The possibility of indi­ erential links has not been onsistent with the premises

the activity of the referen­ activity or strength of the presentations in the mind. ce on a variety of different erential linking function is with a particular entity to

139

be named, and retrieving a word. The more complex operation of the referen­ tial function may be traced, in many different forms, in any natural language context where the speaker must 'find the words' to express experience and the listener must carry out the corresponding decoding process. In previous work, individual differences in basic naming speed were related to differences in verbal performance style in generating a brief spoken monologue describing a personal experience (Bucci and Freedman, 1978). Persons with more rapid naming speeds, characterized as high in Referential Activity (RA) selected more specific incidents, based their narratives on single, elaborated themes and used fewer first person and more third person pronouns. They also showed greater use of certain literary devices associated with vividness and immediacy of language, such as direct quotes and stylistic use of the present tense in describing past events. Low RA subjects tended to select more diffuse and general themes, and focused on general subjective reactions rather than objective and detailed descriptions of specific events. In the same study, high RA subjects were also found to use more speech related gesticulation-hand movements linked to speech rhythm and intona­ tion patterns, which occur to varying degrees in all normal adults, while talking and at no other times. This was interpreted as indicating more active interconnections of motoric functions with verbal representations in generat­ ing speech. The previous findings indicate that greater integration of verbal and non­ verbal systems, as shown in more rapid naming speed, is related to differences in verbal style in more complex representational acts. The current study continues to explore this hypothesis, using new measures and tasks. First we try to show that individual differences in name retrieval time are related to variation in qualities of narrative language associated with representation of imagery. Second, differences in naming speed will be related to style of generating new naming terms. Name retrieval time is assessed using a simple repetitive naming task. The procedure involves a measure of time to name a small set of familiar colors, repeated in random order on a stimulus card; this is corrected for time to read the corresponding words, as a control for personal tempo or general speed. Quality of language style is assessed using a spoken narrative task in which subjects are asked to talk for a brief period about any personal experience, without interruption by the listener. In terms of the dual code model, such narratives might be generated either by retrieving imagery stored in long term memory and translating this into words, or by retrieving material already stored in verbal form. The former process requires an active referential link­ ing function; the latter relies on connections within the verbal system. These

140

W. Bucci

two modes of generating a narrative will lead to corresponding differences in verbal style. To the extent that a description is generated directly through retrieval of nonverbal, imagistic representation, the language is likely to be relatively specific and concrete, with large amounts of objective detail. Nar­ ratives based on material already labeled and categorized and stored in verbal form are likely to be more general and abstract. This formulation is based on the premise of dual coding theory, that the links between verbal and nonverbal systems are most direct for concrete words and the entities they represent, and less direct for abstract concepts and words. Also, the imagistic code has the capability of preserving more specific and concrete detail than is registered in a verbal description, as shown by Shepard (1966), Neisser (1970), Paivio (1969) and others. These differences in style of generating a narrative will be assessed using a system of scaled judgment scores. The prediction is that subjects high in RA will produce narratives judged as rela­ tively specific, definite and concrete. Their narratives will also be more objec­ tive in focus, relying more on reference to other persons, places and things and less on description of subjective personal states; this will be measured in terms of personal pronoun usage. Style of generating naming terms is assessed using a task in which subjects are asked to produce labels for a set of subtly different color tones within the same color families. Specific color terms distinguishing these tones are un­ likely to be available within the subject's vocabulary. Labels might be gener­ ated either by producing metaphoric expressions referring to entities posses­ sing the color, or by combining basic color terms and modifiers. The metaphoric approach relies more heavily on the referential linking function­ the linkage of words to images of entities outside the color domain. The building up of a color description by combining color and color-related terms into a string involves greater use of connections within the verbal system. Method Subjects

The 51 subjects were nursing students at a New York City teaching hospital, who volunteered for the study and received a small payment for participating. Their educational backgrounds ranged from a minimum of two years of col­ lege to undergraduate and even graduate degrees. All were fluent speakers of standard English and within the normal range on a measure of word read­ ing speed.

corresponding differences in 5 generated directly through , the language is likely to be ants of objective detail. Nar­ egorized and stored in verbal :t. This formulation is based te links between verbal and words and the entities they nd words. Also, the imagistic eific and concrete detail than by Shepard (1966), Neisser mces in style of generating a caled judgment scores. The uce narratives judged as rela­ itives will also be more objec­ er persons, places and thin¥s .ates; this will be measured In

using a task in which subjects ifferent color tones within the iguishing these tones are un­ ulary. Labels might be gener­ 1S referring to entities posses­ r terms and modifiers. The referential linking function­ itside the color domain. The color and color-related terms IS within the verbal system.

York City teaching hospital, nall payment for participating. minimum of two years of col­ .ees. All were fluent speakers ,ze on a measure of word read-

'V

Linking words and things

141

Procedure All subjects were seen individually in a single session lasting approximately 45 minutes. The session included the three tasks outlined above as well as several measures used for other research purposes; the latter will not be discussed here. Order of administration was the same for all subjects. The entire session was video and audio-taped for later transcribing and scoring. All measures were administered and scored by trained research assistants unacquainted with the purpose of the study. Naming speed The naming speed measure is based on color naming corrected for word reading times. The word reading task consists of 100 color words, printed in black ink; these were tokens of four types-'red', 'green', 'yellow', 'blue'-re­ peated in random order in a 10 x 10 matrix. The color nanling task consists of strings of patches of the corresponding 100 colors, in the same format. The subject is instructed to read the words; then to name the colors, as quickly as possible. The score for each task is time in seconds for the entire card. The name retrieval measure is derived by subtracting color naming from word reading times. This measure uses the word reading and color naming cards of the Stroop color-word test (Stroop, 1935). In factor analytic studies of the Stroop test, word reading time has been interpreted as a relatively pure measure of the personal tempo factor (Thurstone and Mellinger, 1953). The correction by subtraction may also be understood in terms of a comparison of processes involved in the two tasks, as discussed elsewhere (Bucci and Freedman, 1978). (We do not use the interference measure commonly as­ sociated with the Stroop, in which color words are printed in noncorrespond­ ing color inks, since the naming and reading tasks provide the necessary data in more direct form.) Narrative language The subjects are instructed to speak for five minutes about any interesting personal experience. They are told that the experimenter will be listening but would prefer not to respond, not even to answer questions, until the five minutes are up. The subject is then expected to talk on his own for this period, although listening responses---e.g., nods or smiles-are given where appropriate. The speech sample is transcribed; the typed transcript is then eval­ uated on three qualitative dimensions: Concrete-Abstract, Specific-General, and Definite-Vague, on scales of 1-3, by three judges. The three scales are con­ ceptually distinct;' however they are used here not primarily as reflecting separate characteristics of the narratives, but as different ways of making

142

W. Bucci

inferences to the same underlying dimension, the immediacy of mental imag­ ery as used in the task of generating spontaneous speech. The three scales are combined to yield an overall Immediacy Index. In addition, a direct quantitative measure of Subjectivity vs. Objectivity of the monologue was also used, based on a count of first person versus other personal pronouns. (First person singular pronoun types include 'I, me, my, mine, myself'.) Generating color names Subjects are shown 14 color patches, on a stimulus card, and asked to provide names for each. The colors were selected from within the brown, green and red families, varying as to tint or shade. Subjects are instructed to produce a label for each color that serves to distinguish the color from all the others and that has general and shared rather than idiosyncratic reference. Thus the term 'brown' alone would not be acceptable since it could apply to several of the stimuli; a response such as "the color of my sister's car", or "the color that makes me feel sad", would not be acceptable because the reference is personal rather than shared. Descriptions based on position on the card, e.g., "second from the top on the right", are also not permitted. Any other communicative descriptions reflecting shared or shareable knowl­ edge and experience are acceptable. Once the instructions have been communicated, several attempts may be required before the subject understands them fully and is able to find a suitable term for each item. The task was carried out in several phases, designed to ensure that each subject eventually retrieved or generated an appropriate label for every item. In this paper we are concerned only with . the nature of the names that were produced in the final 'delivery' phase, not with the subject's efforts in generating them. (Data for the early phases of this study, including individual differences in style and facility of generating naming terms, will be discussed in a separate report). In the delivery phase, subjects were instructed to repeat the names as quickly as possible. This instruction was intended to provide a standard context for evaluating the, response, and to eliminate discursiveness; speed of naming per se is not relev­ ant to evaluation of performance on this task. Errors were essentially elimi­ nated in the preparatory procedures and are not scored here. Each response produced in the final delivery phase is classified into one

IThey may also be distinguished empirically, e.g., compare the following sentences: (a) He had something to eat; (b) He had a large breakfast at home; (c) He had steak and eggs with hash brown potatoes, English muffins, and coffee; (d) He had a small juicy steak, with soft scrambled eggs, a buttered English muffin,and all the trimmings. The first sentence is rather vague. The next three are more definite. Sentence (c) is more specific; (d) is more concrete.

Linking words and things

e immediacy of mental imag­ speech. The three scales Index. In addition, a direct ity of the monologue was also rer personal pronouns. (First ny, mine, myself'.) DUS

stimulus card, and asked to eted from within the brown, de. Subjects are instructed to .inguish the color from all the than idiosyncratic reference. ptable since it could apply to color of my sister's car", or >t be acceptable because the riptions based on position on ght", are also not permitted. 19 shared or shareable knowl­ ted, several attempts may be fully and is able to find a irried out in several phases, ly retrieved or generated an . we are concerned only with the final 'delivery' phase, not [Data for the early phases of tyle and facility of generating eport). In the delivery phase, as quickly as possible. This d context for evaluating the l of naming per se is not relev­ Errors were essentially elimi­ rt scored here. y phase is classified into one 1

following sentences: (a) He had something d eggs with hash brown potatoes, English ibled eggs, a buttered English muffin, and :e are more definite. Sentence (c) is more

143

of the following categories: (1) Basic color terms with modifiers, defined as consisting of the primary colors (blue, red, yellow), colors formed by combin­ ing two of these (purple, orange, green), the basic shades (black, brown, white), and shades formed by combining white with the other two (tan, grey). These may be cornbined with each other, or with other modifying terms, e.g., 'a very deep orangey brown'; (2) Specific color terms, defined as any color words other than the basic terms listed above, e.g., maroon, mauve, sienna, either alone, modified as to tint or saturation, or in combination; (3) Metaphoric terms, i.e., descriptive terms drawn from noncolor domains, e.g., 'this color is like dried mud'. Results

Before presenting the findings on the two linguistic performance tests, data for name retrieval times in this sample in comparison to other subject groups will be presented briefly here. As shown in Table 1, the current sample showed a mean of 39.2 seconds for word reading (for the 100 item card) compared to 58.2 seconds for color naming, with a name retrieval score of 19 seconds, computed by subtracting word reading from color naming time. The performance time on each card was approximately five seconds slower than the times for the college student sample reported previously (Bucci and Table 1.

Comparison ofname retrieval scores based on reading and naming speeds in several experimental groups Sample

This study

Number of subjects

51

Mean word reading time (sec.)

Mean color naming time (sec.)

Name retrieval time score (naming minus reading time) (sec.)

39.2 (S.D.

58.2 (S.D. == 8.8)

19.0 (S.D.

53.3 (S.D. == 11.2) 58.24 57.4 61.0

19.3 (S.D. 20.2 18.2 19.1

= 4.8)

= 7.5)

Comparison groups"

Bucci and Freedman, 1978 Jensen, 1965 Broverman, 1960 Comalli et al., 1962

40 436 46 63

34.0 (S.D. 38.09 39.2 41.9

= 5.8)

= 8.6)

*Name retrieval score computed from data published by others is difference between reported group means rather than means of differences; therefore S.D. not available.

144

W. Bucci

Freedman, 1978) and very close to the means of the other three groups. The latency differences among the groups presumably reflect slight differences in overall personal tempo; the name retrieval score is strikingly similar across all groups. The data suggest that a mean name retrieval time of about 20 seconds reflects a kind of constant in the normal adult population. The con­ sistency in name retrieval score, shown in Table 1, is found even in the face of slight differences in the nature of the stimuli and procedures of adminis­ tration of this non-standardized task, in the several different laboratories reported here. Subjects with name retrieval difference scores below the sample mean of 19 (or 1.9 seconds per item), i.e., individuals with relatively rapid naming times, were classified as high in Referential Activity; persons with difference' scores greater than 19 were classified as low RA. Four subjects with scores of 19, i.e., at the mean of this sample, were not included in either group. The high RA group had an average name retrieval score of 13.7 seconds, for the hundred items, compared to 26.8 in the low RA group. The two groups were essentially identical in word reading time, with high RA subjects averag- i ing 39.5 seconds, compared to 39.1 for the low RA group. Thus the differ­ ences between the two groups involved the name retrieval function directly, and were not a function of personal tempo or other factors which might affect all subprocesses. Scores on the two verbal performance measures were then compared for the high and low RA groups.

Qualities of narrative language As predicted, subjects showing high referential activity according to the nam-; ing speed measure produced narratives judged as significantly more specific,' concrete and focused than the low RA group. Their descriptions focused on' the objective contents of experience, while low RA subjects tended to pro­ duce global descriptions of subjective states. The verbal performance mea­ sures for the two groups in the narrative task are given in Table 2. The high and low RA subjects differed significantly on all three performance quality scales, as well as on the combined Immediacy Index (t = 2.31, P < 0.025, one-tailed). They also differedsignificantly on the Subjectivity-Objectivity measure (t = 2.34, P < 0.025, one-tailed), defined as the ratio of first person singular to all personal pronouns. The difference in narrative style between the high and low RA groups may be illustrated by excerpts from two monologues covering the same theme, a recent training experience on the obstetrics ward. The high RA subject de­ scribed a series of incidents, with concrete and specific detail, e.g.:

Linking words and things

other three groups. The lect slight differences in strikingly similar across rieval time of about 20 It population. The con­ I found even in the face procedures of adminis­ II different laboratories

ow the sample mean of relatively rapid naming persons with difference iur subjects with scores eluded in either group. ore of 13.7 seconds, for group. The two groups igh RA subjects averag­ group. Thus the differ­ rieval function directly, ctors which might affect ice measures were then

:yaccording to the nam­ nificantly more specific, descriptions focused on subjects tended to pro­ .rbal performance mea­ en in Table 2. The high ee performance quality { (t = 2.31, P < 0.025, Subjectivity-Objectivity the ratio of first person

and low RA groups may sring the same theme, a he high RA subject de­ fie detail, e.g.:

Table 2.

145

Qualities of narrative language: High versus low referential activity sub- . groups (mean scale scores) Subgroup

Number of subjects

28 HighRA LowRA 19 Pvalue (All onetailed r-tests.)

Immediacy scales

Subjectivity ­ objectivity Overall measure immediacy (1st pers. sing. index as proportion of personal pronouns)

Specificgeneral

Concreteabstract

Focusedvague

1.70 2.06

1.56 2.07

1.65 2.03

4.96 6.16

0.41 0.54

( = 1.88 p <0.05

( = 2.92 p < 0.01

( = 2.11 P <0.025

( = 2.31 p <0.025

t > 2.34 p <0.025

... So I walked into this other room and there's this lady laying there and she's all by herself. There was nobody even there with her from her family or any­ thing. So I went into the room and I was holding her hand. I was teaching like the deep breathing, to try to relieve the pain. And she all of a sudden she says to me, 'Oh my water just ruptured.' And I just-I looked and I had to look really close. It was hard to tell. It wasn't a big gushing; it was just a little trickle. And I said, 'all right, just hold on a second and I'll get the doctor' '"

Compare this with a characteristic excerpt from a low RA subject describing the same training experience: ... It's very interesting seeing women you know going through labor and birth. And I always wondered like if I wanted to have children or not and I don't know. It just posed a lot of questions in my mind because like I don't want to have children but then when I see how happy they are after the birth I thought oh it must be nice. But then when you think about what comes after-all the work and all the sacrifices and responsibility. And I don't think I'm-I could handle it ...

Reliability of judgments Judgments on the qualitative dimensions of style were made rapidly and were highly consistent across judges. The three judges showed perfect agree­ ment or only one deviation of one scale point for 92% of the cases on the scale 'Specific-General'; 88% for the scale 'Concrete-Abstract'; and 94% for the scale 'Focused-Vague', compared to 56% expected by chance. If one assumes independence of scores assigned to any individual, and random assig­

146

T

W. Bucci

nation of scores to each person by the judges, then the probability of this much agreement is well under 10- 6 •2 The scales showed considerable intercorrelation: S-G vs. C-A, r = 0.85; S-G vs. F-V, r = 0.81; C-A vs. F-V, r = 0.75. While the dimensions rep­ resented by the scales were conceptually distinguishable, and the scoring procedure focused on one dimension at a time, judges were apparently react­ ing to a common underlying characteristic of the narrative style in evaluating all three, as had been anticipated. Qualities of color descriptions

The results reported here cover the style of color descriptions produced in the final or delivery phase of this task, as discussed above. Data for the two groups on this measure are shown in Table 3. As predicted, the low RA subjects showed significantly greater reliance on basic color words. This category accounted for 0.51 of their responses comTable 3.

Properties of color descriptions: High versus low referential activity sub­ groups (mean proportions) Subgroup

HighRA LowRA Pvalue (one-tailed t-tests)

Number of subjects

28 19

Content categories Basic color terms with modifiers

Specific color terms

Metaphoric terms

Other

0.32 0.51 t = 2.19 p <0.025

0.34 0.27 NS*

0.29 0.16 t = 1.75 p <0.05

0.05 0.05

Syntactic structure (use of descriptors in sentence form) 0.16 0.51 t = 3.62 p < 0.001 two-tailed'"

* Not significant. ** Two tailed test is used here because difference was not predicted.

2We regard an event as the assignation of scores to a person by the three judges. The probability of maximal disagreement, defined as scores of (1,2,3), (1,1,3), (1,3,3) is 4/9. In our total of 51 trials, we had in the worst case (concrete-abstract) 6 total disagreements. The probability of this happening by chance is calculated as a Bernoulli process (see Hays, 1972) to be 10- 6 . This estimate could be improved well beyond 10- 6 by taking into account the number of trials with complete agreement. The probability of total agreement (scores of (1,1,1), (2,2,2), or (3,3,3» is 1/9. In our total of 51 trials we had in the worst case 16 total agreements (compared to the chance prediction which is: 1/9 of 51 = 6).

pared to 0.32 in the II nanling style of the 10_

r

[

~

.~

The top one is a lij The next one is... a light shade of~ brown And the... ­ green Uh... the . -,' brown too, and that. .. uh.. .is... comes tan, and' a... uh.. .1ight sh

ut

The high RA su metaphoric expre the same colors as • Lime green, b ocado green, C

O.K., the leav khaki, mud, a the cabinet up

Only the differen

< 0.05, one-taile

specialized color

tendency to dey

property drawn

Differences in An intriguing between the '. subjects were f high RA subj three times as data are shoW¢' significant (p seen in the finding was was in fact co instructions

Linking words and things

e probability of this

pared to 0.32 in the high RA group (t = 2.19, p < 0.025, one-tailed). The naming style of the low RA subjects is illustrated in the following response:

r vs. C-A, r = 0.85;

The top one is a light shade of brown. And the next one is a greenish yellow. The next one is... um ... brown, that's what I said. And the next one after that is a light shade of brown. And the one after that is... uh .. .like a reddish brown And the one after that is a very dark brown. After that is a dull green Uh ... the top again is... uh ... a dark brown. I think I had said that was brown too, and the one after that is... uh ... a shade of green. And after that. .. uh .. .is... uh ... also a light shade of brown. After that is green, and ... then comes tan, and then a... red ... a reddish orange color and after that is a... uh .. .light shade of green.

the dimensions rep­ ole, and the scoring ere apparently react­ ve style in evaluating

riptions produced in ve. Data for the two

ly greater reliance on their responses com­

The high RA subjects used more of both specialized color terms and metaphoric expressions, as illustrated in the following two examples, naming the same colors as in the example above (in reverse order): Lime green, burnt orange, flesh, forest green, maple, olive green, maroon, av­ ocado green, charcoal grey, red, beige, mahogany, mustard and chestnut; O.K., the leaves, the color in your blouse, my coat, Christmas-like ... uh ... clay, khaki, mud, a meadow, licorice, brown on the curtains, mud, peanut shell, and the cabinet up there.

referential activity sub-

Other

0.05 0.05

Syntactic structure (use of descriptors in sentence form) 0.16 0.51 t = 3.62 P <0.001 two-tailed**

licted.

ree judges. The probability of

nir total of 51 trials, we had in f this happening by chance is ould be improved well beyond : probability of total agreement : worst case 16 total agreements

147

Only the difference in use of metaphoric terms was significant (t

=

1.75, p

< 0.05, one-tailed). Thus the findings do not demonstrate a more extensive specialized color vocabulary in the high RA subjects. Rather they point to a tendency to develop naming terms by referring to instances of the color property drawn from outside the color domain. Differences in syntactic structure An intriguing result, which had not been predicted, concerned a difference between the two groups in syntactic structure of the responses. Low RA subjects were far more likely to deliver their responses as complete sentences; high RA subjects as lists of labels. The low RA subjects averaged more than three times as many responses in sentence form as the high RA group. These data are shown in the last column of Table 3. The difference was highly significant (p < 0.001, two-tailed)." This aspect of verbal style may also be seen in the contrasting examples from the two groups shown above. This finding was particularly striking in that the use of sentences rather than labels was in fact counterproductive in the delivery phase of this task, in which the instructions were to name the stimuli as fast as possible. Producing complete 3A

two-tailed test was used here since the finding was not predicted.

148

W. Bucci

sentences obviously lengthened response time, and was not necessary to meet the requirements of the task in any way. Reliability of content categories All color descriptions were scored by two independent raters. The correla­ tion for number of expressions for each subject placed into the 'Basic color word' category versus all other categories was 0.96. This is not surprising since the basic color terms included in this category were listed explicitly. The few differences between the raters that did occur reflected either error or occasional disagreement concerning the status of certain unusual modifying terms. Reliability was also high (r = 0.88) for placement of items into the categories of specific color terms versus metaphoric terms, although below that for the simpler distinction discussed above. Differences here largely re­ flected disagreement as to the extent to which certain terms originating out­ side the color domain, e.g., 'cinnamon', or 'mustard' were now to be consid­ ered specific color terms; i.e., the extent to which the color denotation was generally to be accepted as part of the intension of the term. Where an item's status as a color term was questioned by either judge, it was classified as a metaphoric term, i.e., as involving associations outside the color domain, for purposes of this analysis. Discussion

This study has focused on the referential linking function, which involves the system of connections between verbal and nonverbal systems in the mind­ finding words for representations of 'things' in generating language, and link­ ing words back to nonverbal referents in interpreting the language of others. Individuals differing in a simple measure of direct naming speed were found to show related stylistic differences in two more complex verbal tasks, which involve retrieval of stored information and organization of spontaneous speech. The relation of variation in naming speed to performance in more complex linguistic tasks throws a new light on the cognitive significance of this basic referential act. According to our approach, all types of 'naming', the simplest as well as the most complex, are seen as active and significant mental efforts, reflecting the operation of the referential linking function in varying ways. The speaker must enter semantic memory with a nonverbal representation and retrieve lexical items in the form required to direct articulation. This is the case even for the repetitive naming task, in which a small set of stimuli are given and the naming terms are readily available and held in a short-term

T

r

i- f r ~

i~

r

r

"

memory store. Even here; the memory store and to s made anew for each stimu reading, no activation of tl cation of mental imagery , r~ading task. The linkage t ~hff~rent forms of encodin IS given as a written symbo form. In contrast to the rei may be characterized as at is evidence that grapheme part of the process of inden reading aloud actually occu The formulation of direc cognitive function, one tha with certain conventional , as an automatic associative act (Broverman, 1960, Den i.e., if reading is a more cc that naming always takes II the well-known Stroop intei Both the time differentis phenomena observed in m (Dyer, 1973; Jensen and Ro ~o psychologists operating w Ing processes. These effects ential discriminability of the tial training of the two typl other words, these phenome stimulus or response sides ir is argued here that these « mediating connections whicl because the extra step invol The interference effect arise tion must be bypassed in fa' In this study, individual d demonstrated in speed of d performance in two more COl greater integration of verba metaphoric style of the high a tendency to use imagery a

r

Linking words and thines

ot necessary to meet

:raters. The correla­ into the 'Basic color his is not surprising listed explicitly. The cted either error or 1 unusual modifying nt of items into the ms, although below nces here largely re­ .rms originating out­ re now to be consid­ olor denotation was rm. Where an item's it was classified as a he color domain, for

which involves the stems in the mind­ ~ language, and link­ ~ language of others. ig speed were found ~ verbal tasks, which ion of spontaneous

l,

nee in more complex ificance of this basic laming', the simplest leant mental efforts, lon in varying ways. erbal representation articulation. This is l small set of stimuli I held in a short-term

149

memory store. Even here it is necessary to search or scan the set of words in the memory store and to select one; the linkage of percept to word must be made anew for each stimulus as it appears in random order on the card. In reading, no activation of the referential linking system is required. The evo­ cation of mental imagery would in fact tend to interfere with a simple word r~ading task. The linkage takes place within the verbal system only, between different forms of encoding of the same lexical item. The same word which is given as a written symbol is transformed into, or connected to a phonemic form. In contrast to the referential linking function involved in naming, this ~ay .be characterized as at least in part an automatic process. That is, there IS evidence that grapheme to phoneme translation is evoked as an intrinsic part of the process of indentifying written words, in all cases whether or not reading aloud actually occurs (Navon and Shimron, 1981). ' The formulation of direct repetitive naming as incorporating a significant cognitive function, one that is absent in reading words aloud, is at variance with certain conventional views. Naming has frequently been characterized as an automatic associative process and reading as a more complex symbolic act (Broverman, 1960, Denckla and Rudel, 1976). However, in these terms . . ' i.e., If reading is a more complex task, it is difficult to account for the fact that naming always takes longer than reading, and by the same token, for the well-known Stroop interference effect. Both the time differential and the interference effects are highly robust phenomena observed in many studies, with virtually no counterinstances (Dyer, 1973; Jensen and Rohwer, 1966). Both effects have appeared puzzling ~o psychologists operating within the standard views of the naming and read­ mg processes. These effects cannot be accounted for in terms of either differ­ ential discriminability of the two types of stimuli (Fraisse, 1969), or differen­ tial training of the two types of responses (Brown, 1915; Ligon, 1932). In other words, these phenomena are not a function of differences on either the ~timulus or response sides in the reading versus naming processes. Rather, it IS argued here that these effects follow from the different nature of the mediating connections which are required. Naming takes longer than reading beca~se the extra step involving the referential linking function is required. The Interference effect arises because the simpler, more immediate connec­ tion must be bypassed in favor of one which is not so direct. In this study, individual differences in activity of the referential system, as demonstrated in speed of direct naming, have been related to variation in performance in two more complex linguistic acts. Subjects high in RA showed greater integration of verbal with nonverbal modalities in both tasks. The metaphoric style of the high RA subjects in generating color terms reflected a tendency to use imagery as the basis of classification systems. In contrast,

150

W. Bucci

T r

fr the low RA subjects discriminated and characterized the colors through com­ bining strings of property terms. The difference in reliance on verbal connections versus nonverbal referents may also be seen in the finding of differences in syntactic structure between the two RA groups. This finding, as to use of labels alone by the high RA subjects versus sentence forms in the low RA group, raises an interesting question about which we can only speculate at this time. We suggest that for the high RA subjects the entire color description had the grammatical func­ tion of predication; these predicate forms were attributed directly to the color stimuli. The low RA subjects connected their color descriptions to verbal expressions which identified or pointed to each color item, and which functioned grammatically as subject terms, rather than to the color stimulus directly. Thus they generated complete sentence forms. In this sense, we may say that in syntactic structure as well as content, high RA subjects showed a greater capacity to link words directly to percepts, i.e., to concrete nonverbal representations. In their narrative descriptions, as in generating color names, the high RA subjects produced the type of language that suggests a closer link to specific imagery in the speaker's mind, and is likely to evoke it in the listener as well. In the excerpts above, the high RA subject talks about, in effect points to "this lady lying there"; the low RA speaker refers to the general category of "women you know". The high RA subject gives a dramatic description of a birth; the low RA individual, talking about the same training experience, produces a subjectively focused and abstract discussion of the responsibilities and pleasures involved in having children. We would suggest that the long term memory representations of the high RA subjects include more imagery, in a form that is retrievable by words. The same point is made by the pronoun data. High RA subjects tended to describe objects, people and events rather than their own subjective states. Relying on objective material to communi­ cate experience is in effect the literary technique of the objective correlative, the metaphoric approach. The speaker attempts to communicate inner ex­ perience by describing a segment of external reality, in the hope of evoking a corresponding experience in the listener. In the contrasting low RA style, the speaker describes his subjective state in general terms. Assumptions concerning imagery storage and retrieval For purposes of this discussion, only the most general assumptions are required as to the nature or format of imagistic representations in long term memory, and the processes by which they are retrieved and ultimately linked to words. These assumptions are generally compatible with current dual code and imagery formulations (e.g., Bower, 1970; Kosslyn and Pomerantz, 1977;

rr

r

r~

r~

.

Shepard, 1978). Acr in long term memo properties of object stored in long term I may be registered : entered into workin This formulation than its phenomem semblance between image, and the extei plex functions of the tions or reflections ( Interaction between. at each stage of the may be linked to in nonverbal represents of long term memor trieval. For purposes of th if there are two (or postulate a system of differ in activity or SI the manner of their l ability per se. The na connections at the v remains to be explore

Relation to verbal • In previous work, vary independently ( measures and by stall Vocabulary subtests Wechsler, 1955). The RA construct i of imagery ability as (1970), Paivio (1971), a rich inner life, with, being able to translate will investigate this is measures, including b well as certain subte abilities.

Linking words and things

olors through com­ ionverbal referents

structure between by the high RA .ises an interesting We suggest that for grammatical func­ lirectly to the color criptions to verbal item, and which the color stimulus this sense, we may subjects showed a concrete nonverbal ie

ames, the high RA iser link to specific :helistener as well. in effect points to ~eneral category of tic description of a aining experience, the responsibilities ggest that the long ude more imagery, ide by the pronoun ~ and events rather terial to communi­ jective correlative, nunicate inner ex­ te hope of evoking ting low RA style,

i.

11 assumptions are ations in long term d ultimately linked 1 current dual code I Pomerantz, 1977;

151

Shepard, 1978). According to this basic approach, some information is stored in long term memory in imagistic form, in schemata representing objects, properties of objects, and the relations between them. Other information is stored in long term memory in some kind of linguistic format. Stimulus input may be registered in long term memory, and also may be retrieved and entered into working memory, in either code. This formulation focuses on the functional significance of imagery rather than its phenomenological status; and requires no assumption as to re­ semblance between the underlying imagistic representation, the experienced image, and the external physical stimulus (Shepard, 1978). Images are com­ plex functions of the underlying representational forms, not simple projec­ tions or reflections of them (Kosslyn, 1980; Kosslyn and Pomerantz, 1977). Interaction between nonverbal and verbal representational systems is possible at each stage of the encoding, storage and retrieval processes. Verbal input may be linked to imagistic structures at the time of encoding; verbal and nonverbal representations may be stored in connected form in the schemata of long term memory; or new connections may be made at the time of re­ trieval. For purposes of this discussion, the crucial implication of this model is that if there are two (or more) representational codes, then it is necessary to postulate a system of referential connections between them. Individuals may differ in activity or strength of the system of referential connections, and in the manner of their operation, just as they may vary in verbal and imagery ability per se. The nature of the activity of the complex system of referential connections at the various levels or stages of the representational process remains to be explored. Relation to verbal and imagery abilities In previous work, (Bucci and Freedman, 1978), RA has been shown to vary independently of verbal ability as assessed by several verbal fluency measures and by standard intelligence measures such as the Similarities and Vocabulary subtests of the WAIS (Wechsler Adult Intelligence Scale, Wechsler, 1955). The RA construct is also conceptually distinct from the various dimensions of imagery ability as investigated by Galton (1880), and later by Neisser (1970), Paivio (1971), and others. It is at least theoretically possible to have a rich inner life, with vivid imagery and strong 'image to image' links, without being able to translate this into verbal form. A study is now under way which will investigate this issue directly by comparing RA with a varied battery of measures, including both self-report and functional measures of imagery, as well as certain subtests of the WAIS which primarily tap visuo-spatial abilities.

152

W. Bucci

Conclusions and implications

The importance of the referential linking system for cognitive and linguistic functioning, and the impact of individual differences in activity of this system, have not been explicitly considered within the developmental, psychometric or experimental approaches. We suggest that individual differences in RA should be considered along with variation in verbal and imagery ability as these may affect performance on a variety of experimental tasks. These in­ clude the symbolic comparison tasks discussed above; the tasks involving memory for pictures versus memory for words, used by Paivio and others (Paivio and Csapo, 1969); as well as any task in which imagery ability appears to be a factor affecting verbal performance. RA variation may also playa siginificant role in a wide range of domains, including the literary arts, and scientific and mathematical thought. The importance of imagistic reference as underlying abstract, symbolic representation has been noted by many sci­ entists and mathematicians (Hadamard, 1945). While this paper began with a focus on individual differences, our purpose was not primarily to develop a typology, but more basically to broaden our conceptualization of the referential linking function. The emphasis on syntax, i.e., on word to word connections, in recent years, may have tended to divert attention away from the basic referential process, the linkage of word-sym­ bols and thing-representations in the mind. As Stern pointed out more than half a century ago, the most important event in the mental life of the child is the discovery that "every thing has a name" (Stern, 1927, p. 46). The impact of this linking process in the mental life of the adult remains to be explored. References Bower, G.H. (1970) Analysis of a mnemonic device. Amer. Scient. 58,496-510.

Broverman, D.M. (1960) Dimensions of cognitive style. J. Person., 28, 163-185.

Bucci, W. and Freedman, N. (1978) Language and hand: The dimension ofreferential competence. J. Person.,

46, 594-622.

Brown, W. (1915) Practice in associating color names with colors. Psychol. Rev., 22, 45-55. Comalli, P.E., Wapner, S. and Werner, H. (1962) Interference effects of Stroop color-word test in childhood, adulthood and aging. J. genet. Psychol., 100,47-53. Denckla, M. and Rudel, R. (1976) Rapid "automatized" naming of pictured objects colors, letters and numbers by normal children. Neuropsychol., 14, 471-479. Dyer, F.N. (1973) The Stroop phenomenon and its use in the study of perceptual cognitive and response processes. Mem. Cog., 1, 106-120. Ernest, C.H. and Paivio, A. (1969) Imagery ability in paired-associate and incidental learning. Psychon. Sci., 15, 181-182.

~

I

Fraisse, P. (1969) Why is naming Galton, F. (1880) Statistics of mel Hadamard, J. (1945) The psycholc 1954.

Hays, William L. (1973) Statistics, Jensen, A.R. and Rohwer, W.D. Kosslyn, S.M. (1980) Image and A Kosslyn, S.M. and Pomerantz, J.J Cog. Psychol., 9, 52-76. Ligon, E.M. (1932) A genetic stut Navon, D. and Shimron, J. (1981 from Hebrew. J. verb. Lea Neisser, U. (1970) Visual imagery: Boston, Little Brown & Cc Paivio, A. (1969) Mental imagery Paivio, A. (1971) Imagery and Vel Paivio, A. (1978a) Comparisons oj Paivio, A. (1978b) Mental compar Paivio, A. (1982) Individual differ (Eds.) Cognitive Research i Verlag der Wissenschaften Paivio, A. and Csapo, K. (1969) C Shepard, R.N. (1966) Learning and Shepard, R.N. (1978) The mental Stem, W. (1927) The chief periods Child Language: A Book OJ Stroop, J.R. (1935) Studies of inte Thurstone, L.L. and Mellinger, J.J University of North Carolin Wechsler, D. (1955) Manual for th« tion.

Resume

Dans cette etude on etudie les diffe: referentielles entre les representati (RA) mesure par la rapidite de del (1) la production de termes descript breves d'experiences personnelles. qualites d'objectivite et de vivaein termes de couleur c'est a dire qu'il couleurs. Les performances dans ce mot' et Ie groupe de sujets avec un fonction impliquee dans Ie RA est standard verbale et d'agilite verbal«

Linking words and things

~

and linguistic of this system, , psychometric erences in RA .gery ability as isks. These in­ :asks involving vio and others ability appears lay also play a erary arts, and :istic reference d by many sci­

s, our purpose

o broaden our iasis on syntax, snded to divert : of word-sym­ out more than fe of the child 7, p. 46). The remains to be

ipetence, J. Person.,

-55.

srd test in childhood,

, letters and numbers

mitive and response

mingo Psychon. Sci.,

153

Fraisse, P. (1969) Why is naming longer than reading? Acta PsychoI. ,30,96-103.

Galton, F. (1880) Statistics of mental imagery. Mind, 5,301-318.

Hadamard, J. (1945) The psychology of Invention in the Mathematical Field. New York, Dover Publications,

1954. Hays, William L. (1973) Statisticsfor the Social Sciences. New York, Holt, Rinehart & Winston. Jensen, A.R. and Rohwer, W.O. (1966) The Stroop color-word test: A review. Acta Psychol., 25, 36-93. Kosslyn, S.M. (1980) Image and Mind. Cambridge, Mass., Harvard University Press . Kosslyn, S.M. and Pomerantz, J.R. (1977) Imagery, propositions and the form of internal representations, Cog. Psychol., 9, 52-76. Ligon, E.M. (1932) A genetic study of color naming and word reading. A mer. J. Psychol., 44, 103-121. Navon, D. and Shimron, J. (1981) Does word naming involve grapheme to phoneme translation: Evidence from Hebrew. J. verb. Learn. verb. Behav., 20, 97-109. Neisser, U. (1970) Visual imagery as process and as experience. In: Antrobus, J-S. (Ed.) Cognition and Affect. Boston, Little Brown & Co. Paivio, A. (1969) Mental imagery in associative learning and memory. Psycho/. Rev., 76,241-263. Paivio, A. (1971) Imagery and Verbal-Processes, New York, Holt. Paivio, A. (1978a) Comparisons of mental clocks. J. expo Pschol.: Hum. Percep. Perf., 4, 61-71. Paivio, A. (1978b) Mental comparisons involving abstract attributes. Mem. Cog., 6, 199-208. Paivio, A. (1982) Individual differences in coding processes. In F. Klix, J. Hoffman and E. van der Meer (Eds.) Cognitive Research in Psychology: Recent Approaches, Designs, and Results. VEB Deutscher Verlag der Wissenschaften DDR. Paivio, A. and Csapo, K. (1969) Concrete-image and verbal memory codes. J. expo Psycho/. , 80, 279-285. Shepard, R.N. (1966) Learning and recall as organization and search. J. verb. Learn. verb. Behav., 5,201-204. Shepard, R.N. (1978) The mental image. Amer. Psychol., 33, 125-137. Stern, W. (1927) The chief periods of further speech development. In A. Bar-Adon and W. Leopold (Eds.), Child Language: A Book of readings. Englewood Cliffs, NJ, Prentice Hall, 1971. Stroop, J.R. (1935) Studies of interference in serial verbal reactions. J. expo Psychol., 18, 643-662. Thurstone, L.L. and Mellinger, J.J. (1953) The Stroop Test. Chapel Hill, NC, The Psychometric Laboratory, University of North Carolina, No.3. Wechsler, D. (1955) Manual for the Wechsler Adult Intelligence Scale. New York, The Psychological Corpora­ tion.

Resume Dans cette etude on etudie les differences individuelles dans l'activite et la puissance du systeme de connections referentielles entre les representations mentales verbales et non verbales. Le niveau d' Activite Referentielle (RA) rnesure par la rapidite de denomination est reliee aux performances dans deux taches plus complexes: (I) la production de termes descriptifs pour distinguer des couleurs proches; et (2) la production de descriptions breves d'experiences personnelles. Les sujets rapides pour denomer produisent des discours ayant de grandes qualites d'objectivite et de vivacite. lIs utilisent egalement un style plus metaphorique pour indiquer des termes de couleur c'est a dire qu'ils utilisent une terminologie se referant a des entites hors du domaine des couleurs. Les performances dans ces deux taches indiquent des liens entre une utilisation forte de 'I'image au mot' et le groupe de sujets avec un RA eleve, le groupe avec un RA bas s'appuie plus sur le mot a mot. La fonction impliquee dans le RA est dinstinct du type de capacite abstraite mesuree par les tests d'intelligence standard verbale et d'agilite verbale. Elle est aussi distincte de la capacite d'imagerie.