Cognitive Development, 11, 265-294 (1996) Developing Organization of Mental Verbs: Evidence for the Development of a Constructivist Theory of Mind in Middle Childhood Paula J. Schwanenflugel University of Georgia William V. Fabricius Arizona State University Caroline R. Noyes University zyxwvutsrqp Oglethorpe Tw o e xp e rime nts b y e xa m ining me nts, e xa m ine d the d e ve lo p m e nt c ha ng e s in the o rg a niza tio n c hild re n a nd a d ults ra te d the sim ila rity w a y the y fe lt the y use d the ir m ind thirty-fo ur lo -ye a r-o ld s, w e re c la ssifie d c o g nitive a nd 27 a d ults. mo nito rs yo ung e r In Exp e rime nt d isting uishe d p ro c e ssing c hild re n 1 use d thirty-six 2, 9- a nd 11-ye a r-o ld a b ility (Ma rkm a n, w e re 1981). 8- a nd c hild re n Fifte e n use d , a nd 33 a d ults a lso me nta l ve rb s a c c o rd ing to the a sp e c ts o f me nta l a c tivity. O ld e r c hild re n o n the a nd c o m p re he nsio n a sp e c ts o f a c o nstruc tivist c hild ho o d In b o th e xp e ri- sc a ling (MDS) a na lyse s o f e a c h g ro up ’ s ra ting s ind ic a te d p la c e d g re a te r e m p ha sis c hild re n mo nito ring a nd 15 no nm o nito ring p a rtic ip a te d . Multid im e nsio na l tha t p a rtic ip a nts in m id d le o f p a irs o f me nta l ve rb s in te rms o f the in e a c h o ne . Exp e rime nt a c c o rd ing to the ir c o g nitive mo nito ring info rm a tio n o f a the o ry o f m ind o f me nta l ve rb s o f kno w ing . c e rta inty no nmo nito rs. c e rta inty a sp e c ts a nd a nd c o m p re he nsio n a sp e c ts o f me nta l a c tivity It is c o nc lud e d the o ry o f m ind d e ve lo p d uring m id d le tha n tha t imp o rta nt c hild ho o d . Research on the ontogeny and phylogeny of naive theories of mind is important because of the fundamental role that mentalistic explanations The first two authors contributed equally to the work reported here. The research was supported by NICHD Grant No.ROlHD28796. We thank Cam Corley, Kim Ballein, Ian Gilbert, Trisha Howell, Donna Mansell, Stephanie Louthan, Phil McDonald, and Karen Shick for their assistance in running subjects, Joyce Alexander and Kelleigh Bigler for conducting the pilot experiment, and Justin Prost for assistance with coding comprehension monitoring. We thank the students and teachers of Lawrenceville Elementary and the Tempe Public Schools for their participation. Portions of this work were presented at the 1993 meeting of the Psychonomic Society in Washington, DC. Correspondence and requests for reprints should be sent to Paula J. Schwanenflugel, 325 Aderhold Hall, University of Georgia, Athens GA 30602. 265 266 P. J. Schwanenflugel, W. V. Fabricius, and C. R. Noyes play in our understanding of the everyday behavior and experiences of ourselves and others (e.g., Frye & Moore, 1991; Perner, 1991; Wellman, 1990). Most research on developing theories of mind has concentrated on young children’s first understanding of the mind. At the other end of the human developmental spectrum, cognitive anthropologists, psychologists, and philosophers have been interested in the folk theories of mind held by adults (Clark, 1987; D’Andrade, 1987; Keesing, 1987). Somewhat surprisingly, there has been relatively little work on the development of a theory of mind during middle childhood, although this period-roughly between 6 and 12 years of age-is likely to contain important changes as children elaborate and extend their initial understanding of mental states and processes. In our work (Fabricius & Schwanenflugel, 1994; Fabricius, Schwanenflugel,Kyllonen,Barclay,& Denton,l989;Schwanenflugel,Fabricius,&Alexander, 1994; Schwanen~ugel, Fabricius, Noyes, Bigler, & Alexander, 1994), we have found indications that one of the major developments during middle childhood is a growing understanding of the constructivist nature of mental processes. Evidence for this in our first study (Fabricius, et al., 1989) came largely from 8- and lo-year-olds’ difficulty understanding that there were distinctive mental processes involved in the mental activities of comprehension and selective attention, We gave children and adults scenarios depicting common, everyday examples of memory, comprehension, inference, and selective attention. We asked participants to rate the similarity of all pairs of the scenarios in terms of how they felt they would use their mind in each one. The 8-year-olds treated comprehension and selective attention activities as simply seeing or hearing. Ten-year-olds treated comprehension activities as Memory. They saw that both involved memory, but they did not understand the unique, additional features of comprehension. Like 8-year-aids, however, they continued to think of selective attention activities as simply perception. In contrast, adults clearly recognized features that distinguished the mental activities of comprehension and selective attention. In a subsequent study (Schwanenflugel, Fabricius, & Alexander, 1994), we replicated these findings of children’s difficulty with comprehension and selective attention. In addition, we found that even 10”year-olds had difficulty forming a concept of recognition. Children tended to focus either on the memory aspects or on the sensory aspects of recognition activities. Only adults showed a concept of recognition. Finally, only adults made a higher level distinction between external cognitive cuing (which recognition and inference activities shared) and internal cognitive cuing (which recall and comprehension activities shared). Our primary interest in these studies was in children’s underlying naive theory of mental processes that would give rise to these conceptual failures (Murphy & Medin, 1985). We suspected that children’s difficulty in forming Verbs of Knowing 267 concepts of comprehension, selective attention, and recognition might be symptomatic of a larger difficulty they had in understanding that the mind is constructive. Processes such as selective attention, comprehension, and recognition operate during the initial acquisition of information. By overlooking these processes, lo-year-olds appeared to believe that one simply came to know information without such intervening processes. In contrast, they were quite aware of distinctive processes that occur later in the information processing flow, including planning, comparison, recall, and inference. These processes reflect how information is manipulated after it is initially acquired. Children’s failure to understand the pervasive role of mental processes during information acquisition would seem to preclude a constructivist view of the mind in which mental processes necessarily interpret, organize, and transform information. In Schwanenflugel, Fabricius, Noyes, et al. (1994) we explored adults’ theory of mind more completely in order to look for more direct evidence that adults viewed mental processes as constructive. We took the approach of studying how adults organized 30 mental verbs of knowing representing a wide range of mental processes. Much research from cognitive anthropology has made it clear that the organization and availability of nomenclature in a domain can provide important information regarding how cultures and subgroups think about a particular domain (Berlin, 1992; Blount & Schwanenflugel, 1993). We measured intensional relations among mental verbs by asking adults to rate the similarity of pairs of verbs and we measured extensional relations by asking adults to apply the verbs to scenarios. We subsequently obtained ratings of the verbs on various attributes from an independent group of adults in order to describe the dimensions and clusters obtained in the simildrity judgment and extension data. MDS analyses of both types of data revealed two dimensions (information processing and certainty). First, adults, organized mental verbs along an information processing continuum ranging from input functions (e.g., notice, observe), through processing and memory functions (e.g., question, learn) to output functions (e.g., zyxwvutsrqponm decide, explain). Verbs at the input end of this dimension were also rated as relatively external and perceptual compared to the internal and conceptual nature of verbs in the center and output end. Second, adults simultaneously organized verbs according to the degree of uncertainty that pervades this information processing system. This dimension was anchored at the highly certain end by memory verbs such as know and learn, and ranged through moderately uncertain processing, input and output verbs such as think, observe, and plan, to highly uncertain processing verbs such as guess and question. The verbs toward the uncertain end were also rated as relying more heavily on inference and requiring information less readily available for the process than other verbs. 268 P. j. Schwanenflugel, W. V. Fabricius, and C. R. Noyes These findings extend our earlier findings that adults view menta’t processes as intervening even during information acquisition, We can now see that this is just one part of their larger view that (a) mental processes form an information processing flow extending from initial information acquisition on through memory and processing to output functions, and (b) mental processes are inherently uncertain to varying degrees. This larger view of the pervasiveness and uncertainty of mental processes suggests two reasons why adults do not view knowledge as a direct copy of information, and provides two cornerstones of adults’ constructivist theory of mind. In the studies presented here, we conducted a modified version of the Schwanenflugel, Fabricius, Noyes, et al. (1994) task to examine the development of the dimensions of certainty and information processing in the organization of mental verbs during middle childhood. Some research using other linguistic methods suggests that children come to incorporate certainty and information processing features into their understanding of mental processes during this period. For example, Moore and his colleagues (Moore, Bryant, & Furrow, 1989; Moore & Furrow, 1991) found that the ability to distinguish three levels of certainty expressed by the mental verbs know, think, and guess appeared sometime between 8 years of age and adulthood. Wing and Scholnick (1986) noted that 11-year-olds (and even 16year-olds) still had difficulty interpreting sentences employing various verbs of uncertainty (e.g., wonder, guess), particularly in cases where there were two potential meanings for the sentence. Finally, Hall and his colleagues (Booth 62 Hall, 1994; Frank & Hall, 1991; Hail & Nagy, 1986; Hall, Scholnick, & Hughes, 1987) proposed that cognitive words and their various meanings are acquired according to a levels of processing hierarchy ranging from words referring to perceptual activities to words referring to evaluation and metacognitive activities In their research, they have found that preschoolers use the evaluation level (consisting of words referring to presuppositions of truth, evaluation of truth, and uncertain states) less than adults do in their spontaneous speech (Frank & Hall, 1991; Hall et al., 1987) and that even lo-year-olds have some difficulty identifying synonyms for the evaluation meanings of polysemous cognitive verbs (Booth & Hall, 1994). In the experiments cited in this article, children and adults judged the similarity of pairs of mental verbs in terms of the way they would use their mind in each one. To make the task manageable for children, we selected a different subset of the 30 verbs used by Schwanenflugel, Fabricius, Noyes, et al. (1994) for each experiment. Only five verbs (memorizing, understanding, knowing, thinking, guessing) were common to both studies, thus allowing us to assess the generalizability of the findings to substantially different sets of verbs. In both experiments, we were interested in assessing the degree to which children’s organization of mental verbs and theory of mind resembled 269 Verbs of Knowing that of adults. In particular we looked essing dimension characterized by a their input/output, external/internal, and a certainty dimension characterized of degrees of certainty, inference, and for evidence of an information zyxwvutsrqponml procdistinction among verbs in terms of and perceptual/conceptual relations, by distinction among verbs in terms information availability. EXPERIMENT 1 Method Participants. Participants were thirty-six 8-year-olds (M = 8;1, SD = = 7;4-9;2), thirty-four lo-year-olds (M = 10;4, SD = $0, range = 9;7-11;2), and 27 adults. The children were from a public school in suburban Atlanta and the majority were white. The adults were students from an introductory educational psychology class who volunteered to participate. $0, range Stimuli and procedure. Stimuli were a representative sample of the 13 mental verbs selected from the 30 used by Schwanenflugel, Fabricius, Noyes, et al. (1994): checking, deciding, discovering, explaining, guessing, hearing, inventing, knowing, memorizing, noticing, pay ing attention, thinking, and understanding. Verbs from all four quadrants of the similarity rating MDS solution from Schwanenflugel, Fabricius, Noyes, et al. (1994) were represented. All verbs from the earlier study had been identified as being representative ways of knowing or coming to know something. The verbs had a frequency of at least 10 or greater per million in the third-grade corpus according to Carroll, Davies, and Richman’s (1971) W ord Frequency Book (M verb frequency = 447 per million, SD = 628). Thus, it is reasonable to assume the children had been exposed to the verbs. For the similarity judgment task, all possible pairs of verbs were created so that participants performed similarity ratings on 78 verb pairs. Approximately 20 verb pairs appeared on a page and the ordering of the pages was counterbalanced across subjects. The children were run singly in a room at their school. The experimenter first presented the children with the list of verbs to be used in the experiment and then provided an example of a situation in which they would use their mind in that fashion (see Appendix). All participants professed to be familiar with the meaning of each verb. The Children were instructed to rate the pairs of verbs on how alike or how different the words were based on the way they would use their mind in each one. They were told not to think about what their body would be doing, but to think about what their mind would be doing. They were asked to rate the pairs of words on a scale of 1 to 7 with 7 meaning that they would use their mind in completely the same way and 1 meaning that they would 270 P. J. Schwanenflugel, W. V. Fabricius, and C. R. Noyes use their mind in completely different ways. They were told to use the other ratings to represent different degrees of using their mind in same and different ways such that, for example, if they would use their mind a little the same and a little differently in the two words, they would give the word a rating of 4. They were encouraged to use the full range of the scale. The participants were then given two practice pairs for verbs not included in the study. They were asked if they had any further questions and were reminded that they should rate the words on what their mind was doing, not what their body was doing. A visual number line was presented with the end points labeled “completely same” and “completely different,” above which increasing bars were represented showing increasing similarity. The experimenter read each pair to the children one at a time and the children gave their rating. The experiment took the children about 20 mins to complete. The adults were given written instructions and completed the rating task on their own. Results MDS procedures were used to uncover the relations participants perceived among the verbs. In MDS procedures, similarity among items is indicated by their spatial proximity in a multidimensional space. Developmental changes in the organization of items and weightings of dimensions in the MDS space can be indicative of fundamental changes in the way that children view the domain (Miller, 1987). MDS typically rakes a group similarity matrix as input. To create a similarity matrix for the similarity judgment task, ratings for each of the 78 comparisons were averaged across subjects at each age. Prior to averaging, however, it was important to determine that there was some agreement in ratings across subjects at each age to ensure generalizability across members of the group. The Cronbach coefficient alpha measure of interrater reliability was calculated from each cohort’s data set. Because this formula usually produces a reliability lower than standard split-half reliabilities, it represents a conservative estimate of the actual interrater reliability and can be thought of as providing a minimum reliability estimate (Borg & Gall, 1983, p. 285). These reliabilities suggested substantial interrater agreement among participants in their judgments regarding the verb pairs (adults, .95; lo-yearolds, .79; &year-olds, 66). Thus, the organization obtained in the group MDSs’ should be reasonably generalizable to the individual participants in each group. Therefore, ratings were averaged across subjects at each age group to form a similarity matrix for each age group. Prior to examining developmental changes in the relative importance of various features in the organization of mental verbs, it is first important to assess whether common structural elements might be identified in the space of all three groups. One way to test for commonality of structure would be Verbs of Knowing 271 to correlate each group’s similarity matrices with the other groups’ matrices. This analysis yielded substantial correlations-.75 between the adults and lo-year-olds; .72 between lo-year-olds and 8-year-olds, .65 between adults and S-year-olds. A second way to examine for commonality of structure is to determine whether each group’s MDS might be characterized by the same dimensions-that is, whether the groups’ MDS’s might share a common interpretation. We performed a separate MDS analysis for each group using the Alscal program of SAS (SAS Supplemental Library Users Guide, 1980). When using multidimensional scaling, the choice of number of dimensions is based on several criteria, among them goodness-of-fit, the number of items, the stability of the solution, and interpretive criteria (Kruskal & Wish, 1978; Shepard, 1972; Shoben & Ross, 1987). We have presented the two-dimensional solution in each case because the number of verbs was small for the three-dimensional solution; the first two dimensions remained stable when going from two to three dimensions; and a common interpretation (see later) seemed reasonable from the solutions generated. For adults, the two-dimensional solution accounted for 79.8% of the variance in the ratings with a stress of .201. For lo-year-olds, the two-dimensional MDS accounted for 73.7% of the variance in the ratings (stress = .2X). For 8-year-olds, the two-dimensional MDS accounted for 61.4% of the variance in the ratings (stress = .265). A descriptive account of the MDS space involves orienting and interpreting the dimensions. This is usually provided by developing interpretations of the space and having either experts or naive informants verify the validity of the interpretations. Alternatively, theory or prior research can enable us to make predictions regarding the potential organization of the space. Because our earlier research (Schwanenflugel, Fabricius, Noyes, et al. 1994) had already indicated a strong candidate for the organization of this space, we correlated the relevant attribute ratings obtained in that study with the position of verbs along potential orientations of the dimension, looking for the best correlation with the dimension. Because the orientation of dimensions is arbitrary in MDS, we examined a horizontal and vertical orientation as well as a 45” rotation for each dimension clockwise and counterclockwise. To test for the zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA information processing dimension, we tested the following attribute ratings: 1. information processing scale in which verbs were rated along an information processing continuum from input to output (1 = input, 7 = output); 2. perceptuaVconceptua1 scale in which verbs were rated according to their perceptual vs. conceptual nature (1 = perceptual, 7 = conceptual); and P. J. Schwanenflugel, W. V. Fabricius, and C. R, Noyes 272 3. exr~~~~~~tern~~ scale in which verbs were rated according to the de- zyxwvutsrq gree to which they referred to the external word or the intern31 world (1 = external, 7 = internal). Normative ratings on these scales for this set of verbs were intercorreiated nonformation pracessing-interna~/external r = .5SI internaiiexternal-perceptual/conceptual r = .94, information processing-perceptual/conceptual r = .65,p < .OS). To test for the certdnty dimension, we used the following attribute r3tings: 4. certainty scale in which verbs were rated according to the degree to which they involve knowing something for certain (1 = certain, 7 = ~~~er~ai~}; 5. inference scale in which verbs were rated according to the degree to which they invofve inferencing (I = low inference, 7 = high inference); and 6. information availability scale in which verbs were rated according to the degree to which they involve information that is readily available (1 = information available, 7 = information not available). Again, normative ratings on these scales for t,hese verbs were all positively intercorreIated (certainty-inference r = .52, certainty-information availabiIity r = .73, information availability-inference r = Sl,p < .05). Table 1 presents the correlations between the various information processing and certainty attributes and the verb positions along the two dimensions for the adults, lo-, and S-year-olds. As seen in Table 1, the organization of mental verbs demonstrated by adults in this study closety resembled the findings of ~~hwanen~uge~, Fabric&s, Noyes, et al., (1994). The First dintension was certainty. This dimension correlated welt with the certainty, inference, and information availability scales, all p < .0.5.1This characterization also seemed to be a reasonable one for both of the children’s solutions. The organization of verbs along the first dimension for both child groups correlated with at ieast two out of three of the attribute rating scales for the certainty attributes, The second dimension for adults reflected a distinction along information processing Lines.As can be seen in Table 1, this dimension correlated reasonably weIt with the information processing, per~eptua~~~o~~e~tual~ and external/in~erna~ scales (ail p < .OS).The organization of items along the second l~~o~~~o~f,a one-tailed test of significance is used for these attribute correfations and other tests of significance whenever the test entails directional hypotheses developed on the basis of our earlier research (Schwanenflugel, Fabricius, Noyes, et al. 1994). 273 zyxwvutsrqp Verbs of Knowing Table 1. Correlations of Attribute Ratings with Position of Verb for Two Dimensional MDS for each Age Group of Experiment 1 Dimension Certainty Inference Information Availability Information Processing Perceptual/ Conceptual External/ Internal Adults Dl D2 [.83 .lO .59 .29 .76] .02 .26 [.73 -.27 .65 - .37 .54] IO-year-olds Dl D2 [.83 .lO .64 .40 .72] .14 -.31 f.40 .22 .64 .30 .54] B-year-olds Dl 1.63 44 .56] - .23 .21 .17 .lO .56 1.52 .51 .38] D2 Note: rcrit = .48, one-tailed -.38 p < .05. dimension of both of the children’s solutions also correlated with these attributes for two out of three of the scales.2 Finally, a third way of testing for commonality of structure is to correlate the verb coordinates on each dimension obtained for each individual group MDS. These correlations can be found in Table 2. As can be seen, coordinates of the items along the certainty dimension showed considerable relationship between each group. However, for the zyxwvutsrqponmlkjihgfedcbaZYXW information processing dimension, there was substantial agreement among the two child groups, but more limited agreement between the child groups and the adult group. In sum, there seemed to be considerable agreement between the children and adults in their organization of these verbs, particularly for the distinctions between mental verbs regarding certainty. Evidence suggested more caution in assuming that there was commonality of structure across groups regarding a distinction along information-processing lines. Combined analysis. Because there was commonality of structure underlying the MDS organization of each group at least regarding the certainty dimension, it was reasonable to assess whether there were developmental differences regarding the weighting or importance attached to this dimension. To do this, we conducted a combined individual differences multidimensional scaling analysis called INDSCAL, in which each age group was entered as one “individual.” The INDSCAL model assumes that the groups This interpretation somewhat better-fitting ,172). also remains a good explanation of the first two dimensions of the three dimensional solution for the &year-old data (‘2 = ,721, stress = P. J. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONML Schwanenflugel, W. V. Fabricius, and C. R. Noyes 274 Table Group’s 2. Correlations Between the Dimension Coordinates Between Each MDS Adults Group Certainty Adults lo-year-olds S-year olds Information Adults lo-year-olds 8-year-olds lo-year-olds S-year-olds 1.00 .97 .71 1.00 .68 1.00 1.00 .41 .53 1.00 .81 1.00 Processing Note: rcri, = .48, one-tailed; p < .OS share a common structure, differing only in the emphasis that they place on the dimensions (Kruskal & Wish, 1978). The two-dimensional INDSCAL solution for this combined data set accounted for 59.2% of the variance in the ratings (stress = .278). The two-dimensional INDSCAL is presented in Figure 1. As expected, the first dimension to emerge from this analysis was related to the certainty aspects of mental activities. Each of the attributes related to certainty correlated highly with the position of items along this dimension: certainty, -84; inference, .64; and information availability, .77, zyxwvutsrqponmlkjihgfed p‘s < .05. A multiple correlation analysis regressing these three attribute ratings together on the position of items along this dimension yielded R = .87, F(3, 12) = 9.36 @ = .05), accounting for 75.7% of the variance in the organization of verbs along this dimension. The second dimension was related to the information-processing aspects of mental activities. This dimension was correlated with the information processing, external/internal, and, especially, perceptual/conceptual scales (.48, .63, and .77, respectively, p < .05). Taken together, these attribute ratings correlated .74 with the position of items along the second dimension, F (3, 12) = 3.57, p = .03, and accounted for 54.4% of the variance in the organization of this dimension. As shown in Table 3, the first dimension was weighted more highly by the adults than the lo-year-olds and, especially, the 8-year-olds, showing a developmental trend of increasing importance placed on these certainty aspects of mental activities. The information-processing dimension appears to show decreasing weight with development, but because this dimension had shown less converging evidence of commonality of structure across ages, we do not wish to place any emphasis on this finding. However, another way to think of the weightings obtained from the INDSCAL would be to compare the relative weightings within each group for the certainty and information-process- 275 Verbs of Knowing Output/ Conceptual/ Internal zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIH II decide p understand n know II diPCOV~ Certain/ Noninferential/ D think Uncertain/ Inferential/ ’ memOnze Available n finavailable check q 9 guess attend @heal PI notice Inpitt Perceptual! External zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPO dimension Figure 1. Certainty dimension (horizontal) and Information-Processing (vertical) from the two-dimensional combined INDSCAL representation of the verb simiIarity judgments from Experiment 1. ing dimensions. Examining the weightings in this way, it can be noted that, with age, the balance of relative weights shifts from the information-processing dimension toward the certainty dimension such that adults clearfy put more emphasis on the certainty aspects of the verbs in their ratings than on the information-processing aspects. Either way that the INDSCAL weightings are compared, it is clear that the developmental trends is toward placing more emphasis on the certainty aspects of mental processes. One potential issue for concern in interpreting the results is the possibility that the &year-olds may have had limited knowledge of a few of the items used. Most of the items in this study were of very high frequency ( > 100 per million) and are likely to be known at least somewhat by all the children. However, three of the items zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONM (~~venii~g, ~e~or~zi~g, and pay ing attention) were of only moderate frequency (ranging from lo-50 per million), so the &year-old children may have had more limited exposure to these items. With P. J. Schwanenflugel, 276 W. V. Fabricius, and C. R. Noyes zyxwvutsrqpo Table 3. Group Dimension Weights for the Two Dimensional INDSCAL Analvsis for ExDeriment 1 Group Adults lo-year-olds 8-year-olds Certainty Processing .74 .63 .44 .30 SO ss MDS, items of similar meaning should appear near each other in multidimensional space. One way to assess S-year-olds’ relative understanding of these items is to determine whether their group MDS placed these items with similar neighbors as the adult MDS did. Examination of the &year-old MDS indicated that their placement of ~nve~~~~g, ~e~?z#r~zing, and paying attent~~n was indeed similar to that of adults. For these verbs, at least two out of three of the nearest neighbors on the adult MDS were present as nearest neighbors on the &year-old MDS. Therefore, it appears that the S-year-olds’ understanding of these potentialIy more difficult items reasonably approximated the understanding possessed by adults. EXPERIMENT 2 Experiment 1 indicated that one of the major developments in the organization of verbs of knowing is an increasing emphasis on the certainty aspects of mental verbs with age. Although important in itself, we know nothing about the relationship of this developing emphasis to actual cognitive behavior. In this study. we were interested in determining the potential relation between the organization of mental verbs along the certainty dimension and the specific metacognitive behavior of comprehension monitoring. Children’s awareness that there are degrees of uncertainty in mental processes should be related to their awareness of the constructivist distinction between what is said in a text and the interpretation one makes in comprehending it. Thus, this study was designed to both replicate the findings of Experiment 1 and to relate the developing organization to a specific metacognitive behavior. Some research has linked understanding of a few selected mental verbs to comprehension in children. Olson and Torrance (1987) found that third graders who were better able to distinguish between zyxwvutsrqponmlkjihgfedcbaZY know, think, and pretend were also more likely to draw inferences in a listening comprehension task. However, it is not clear that children based their distinctions on the certainty aspects of these verbs. In their task, children were asked things like “You (know, think, pretend) that your book is lost. . . Is it really lost?” The phrase “think that” can mean either to hold a belief that you are uncertain about, or to believe with certainty something that is false. Thus, children may Verbs of Knowing 277 have correctly distinguished between know and think either by understanding that two different degrees of subjectively experienced certainty were implied, or by recognizing the linguistic property of factivity-that is, that know implies the truth of the subordinate clause and think does not (Kiparsky & Kiparsky, 1970). In contrast to tasks that focus on only a few mental verbs, our verb similarity judgment task should provide a more sensitive measure of children’s awareness that mental processes vary in certainty for several reasons. First, the similarity judgment task can capture the organization of many different mental verbs representing a wide range of different mental processes, reducing the chance of capitalizing on particular features of a few verbs that might be confused with certainty. Second, the similarity judgment task provides a sensitive measure of the weight given to the certainty dimension relative to other organizational features of mental verbs. In the experiment presented here, we hypothesized that children who are more likely to monitor their comprehension would weight the certainty aspects of mental activity verbs more than children who are less likely to monitor. Children were given the Markman (1981) comprehension monitoring task in which they were read passages containing inconsistencies and then asked questions to ascertain whether the children had detected the inconsistencies. Children were also asked to rate the similarity of a substantially different set of mental verbs from Schwanenflugel, Fabricius, Noyes, et al. (1994) than in the first study. Although the two tasks appear to be quite different, we predicted they would tap metacognitive behaviors and aspects of conceptual structure which eminate from the same developing constructivist theory of mind. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDC Method Participants. Participants were thirty 9-year-olds (M = 9;2, SD = 4;0, range = 8;6-9;10), thirty ll-year-olds (M = 11;2, SD = 5;0, range = 10;7-12;1), and 33 college students. The children were sampled until there were 15 comprehension monitors and 15 comprehension nonmonitors from each age group. The comprehension monitors and nonmonitors were similar in age (9-year-olds: monitor M = 9;2 and nonmonitors M = 9;3; ll-year-olds: monitor M = 11;3 and nonmonitors M = 11;l). The children were from an elementary school in the Tempe school district, and adults were from an introductory psychology class and participated as part of a course requirement. Stimuli. A representative sample of 17 mental verbs of knowing were selected from all four quadrants of the Schwanenflugel, Fabricius, Noyes, et al. (1994) similarity rating MDS: comparing, estimating, examining, explou- P. J. Schwanenflugel, W. V. Fabricius, and C. R. Noyes zyxwvutsrq 278 ing, guessing, knowing, learning, memorizing, obse~ing, questioning, reuson- and underFive verbs (memorizing, understanding, knowing, thinking, guessing) were common to both Experiments 1 and 2. According to Carroll, Davies, and Richman (1971), the mean verb frequency across verbs in the study was 481 per million with an standard deviation of 807, but all verbs had a frequency of at least 10 per million. Thus, it is likely that children had reasonably high exposure to the words used in this study. Children were presented with all possible pairs of verbs so that they rated 136 verb pairs. ing, recognizing, remembering, searching, seeing, thinking, standing. procedure. The procedure for this study was similar to the previous experiment with a few exceptions: For children, stimuli were presented on an IBM computer with a Goldstar Microtouch monitor. Each participant received the 136 pairs in a different order. During the task, the first (randomly selected) verb of the pair was printed at the top of the screen and was simultaneously “read” to the child by the computer using a prerecorded female voice, followed by the second verb. At the bottom of the screen appeared a number line from 1 to 7 with endpoints labeled as “completely different” and “completely the same,” respectively, and a series of increasing vertical lines above the numbers to serve as a visual aid for the participants’ ratings. To make their ratings, children touched the scale on the monitor which highlighted the number touched. Participants could change their mind by touching another number on the scale and, when satisfied with their rating, touched a sentence on the bottom of the screen that said, “To move to the next pair touch here.” Instructions were similar to the previous study. Children were presented with the list of verbs to be used in the experiment and then were provided with an example situation in which they might use their mind in that way. For the five verbs in common with Experiment 1, we provided different examples than in the first study to reduce the potential influence that particular examples might have. Further, children were asked to provide their own examples of each mental verb used in this study. The ~mprehension monitoring task was given to each child about midway through the rating task. The child was asked if he or she wished to take a short break to listen to some stories. The comprehension monitoring task was the explicit condition of Markman’s Study 1 (1981) in which children were asked to help the experimenter write stories so that other children could understand them. The children were read three short stories about 10 sentences long containing obvious inconsistencies. An example of an explicit inconsistency from one story is as follows: Fish must have light in order to see. There is absolutely no light at the bottom of the ocean. It is pitch black down there. When it is that dark the fish cannot Verbs of Knowing 279 see anything. They cannot even see colors. Some fish that live at the bottom of the ocean can see the color of their food; that is how they know what to eat. After they were read each story, children were asked questions to determine whether they monitored their comprehension of the story: (a) What do you think?; (b) Do you have any questions?; (c) Did I forget to tell you anything? and (d) Did everything make sense? Then, if necessary, they were asked other questions relating to the specific inconsistency in the story. We determined, for each story, whether the child gave a clearly adequate indication that he or she detected the inconsistency in the story at any point in response to questions (a)-(d). If they answered any of these questions correctly, the experimenter exited the question sequence and identified the child as monitoring comprehension for that story. Four types of responses were considered clearly adequate: (1) statements that explicitly pointed out both sides of the contradiction (e.g., “You said it’s dark and they can see.“); (2) simple questions about one of the contradictory statements (e.g., “How can the fish see?“); (3) complex questions that referred to both sides of the contradiction (e.g., “If it’s dark, how can they see?“); and (4) causal statements that pointed out the logical consequence of one of the contradictory statements (e.g., “Fish can’t see when it’s dark”). Interrater reliability between two judges scoring all the stories was 98%. Disagreements were resolved by discussion. Children were then classified as comprehension monitors if they gave a clearly adequate indication that they detected the contradiction in at least two of the three stories. Adults were tested in small groups of two to four at a time using paper and pencil. Pairs were presented two orders, one the reverse of the other. The adults were not given the comprehension monitoring task. zyxwvutsrqponmlkjihgfed Results Similarity Judgments. Reliabilities indicated a great deal of commonality among participants of each age regarding the similarity of the verb pairs. Cronbach coefficient alphas were .77 and .68 for the 9-year-old monitors and nonmonitors, respectively, .72 and .78 for the 11-year-old monitors and nonmonitors, respectively, and .93 for the adults. Therefore, we averaged across subject ratings to form the similarity matrix for each group. Before entering all groups into an overall INDSCAL analysis, we assessed whether common structural elements could be identified in the organization of verbs for all the groups. We used the same methods as in the previous experiment. First, we correlated each group’s similarity matrix with the other group matrices. Table 4 shows a significant correspondence between the matrices of all groups. P. J. Schwanenflugel, W. V. Fabricius, and C. R. Noyes zyxwvutsrqp 280 Table 4. Co~ eiations Between SinliIarity Matrices for Each Group in Experiment 2 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFE 11-Year-Olds A ge Group Adult 11 Monitors 11 Nonmonitors 9 Monitors 9 Nonmonitors A dult 1.00 .76 .69 69 .57 Monitors Nonmonitors 1.00 .77 .79 ‘66 1.00 .77 67 Y-Year-Old5 Monitors 1.00 .66 Nonmonitors 1.00 NOW r,,i, = .17, p < .05, two-tailed. Second, to assess for a common interpretation, we performed a separate MDS analysis for each group using the Alscal program of SAS (SAS Supplemental Library Users Guide, 1980). In each case, we present the threedimensional solution because the three-dimensional INDSCAL provided a much better fit for all groups as a whole and, this time, there were enough verbs to merit a three-dimensional solution. However, the interpretations of the first two dimensions remained stable when going from two to three dimensions for each group’s data. For adults, the three-dimensional solution accounted for 87.5% of the variance (stress = .139). For the 11-year-olds, the three-dimensional solution accounted for 83.8% and 81.9% of the variance in the ratings for monitors and nonmonitors, respectively (stress = .136 and .136). For the g-year-old comprehension monitors, the three-dimensional solution accounted for 75.8% of the variance in the ratings (stress = .162) and, for the nonmonitors, 74.0% of the variance in the ratings (stress = .173). As before, we examined these MDS solutions for evidence of the certainty and in~~~~at~~~ processing dimensions using the certaintyand information processing-related attribute ratings from Schwanenflugel, Fabricius, Noyes, et al. (1994) for this subset of verbs as predictors of verb position along the first and second dimensions of each MDS solution. 7be attribute ratings for this set of verbs correiated .78 between certainty and information availability, SO between certainty and inference, and .50 between the inference and information availability scales. The attribute ratings for this set of verbs correlated -67 between the information processing and perceptual/con~eptuaI, .53 between the information processing and externaifinternal, and .89 for the per~eptual/conceptual and external/internal scales. The correlations with the attribute ratings for the certainty and information processing dimension for each group’s data are presented in Table 5. As can be seen in Table 5, the first dimension of the adult MDS sotution could be described by a distinction among verbs along certainty lines. This dimension correlated significantly with the certainty, inference, and infor- 281 Verbs of Knowing Table 5. Correlations of Attribute Ratings With Position of Verb for Three Dimensional MDS for Each Group of Experiment 2 Dimension Certainty Inference Information Availability Information Processing Perceptual/ Conceptual External Internal Adults zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Dl 1.96 D2 -.lO II-year-old .55 .75] .08 .60 .05 [.65 Dl [.91 .44 .62] .16 .52 .16 I.88 Sl .61] D2 .25 .52 .26 1.87 .38 .52] D2 .20 .67 .17 Dl D2 .61] .05 [.76 -.24 .71 -.53 .54] -.Ol r.84 -.27 .78 -.50 .62] monitors Dl 9-year-old - .43 nonmonitors Dl 9-year-old .69 monitors D2 II-year-old -.21 -.60 ,:: .73 -.34 .57] nonmonitors [.78 -.20 .72 .24 .52] -.05 .24 [.72 -.21 .78 .03 .80] No:e: rcrit = .41, one-tailed p < .OS. mation availability scales,p < .05. This characterization also seemed to be a reasonable one for each of the child group solutions for the first dimension. At least two of the three attribute rating scales we associate with certainty correlated significantly with the position of verbs along the first dimension for all child groups. Thus, it can be reasonably argued that the adult and child groups share a common interpretation for this dimension. The second dimension of the adult MDS could be characterized as information processing, as in our first experiment. As can be seen in Table 5, this placement of verbs along this dimension correlated moderately with the information processing, perceptual/conceptual, and external/internal scales, p < .05. Further, this distinction appeared to fit the children’s data as well. The ordering of items along second dimension for each group correlated significantly with each of these related scales. Thus, in this experiment, it seemed clear that the children and adults could reasonably share a common interpretation for this dimension. We had no expectations with regards to the characterization of Dimension 3. Thus, we correlated the ordering of items along Dimension 3 with the attribute ratings on other potential attributes from Schwanenflugel, Fabricius, Noyes, et al. (1994). The ordering of items along Dimension 3 for adults appeared to be related to a distinction between verbs involved in scientific reasoning (e.g., reason, question) and verbs not seen as involved in 282 P. J. Schwanenflugel, W. V. Fabricius, and C. R. Noyes scientific reasoning (e.g., zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIH memorize, guess), and correlated best with the scientific reasoning scale (.60). For the 11-year-olds, the third dimension correlated best with related scales: (a) the amount of verbal information scale (.73) for the monitors, and (b) the visual/spatial-auditory/verbal scale (.48) for the nonmonitors. For the 8-year-olds, the position of items on the third dimension correlated best with the retrieval-acquisition of information scale for the monitors (r = .76) and the visual/spatial-auditory/verbal scale for the nonmonitors (r = S3). Thus, commonality of structure could not be assumed for this third dimension. Finally, we tested for commonality of structure a third way by correlating verb coordinates on each dimension for each groups’ MDS. As can be seen in Table 6, coordinates of items along both the certainty and information processing dimensions showed considerable correspondence between groups, suggesting commonality of structure for these two dimensions. The correspondence between groups was less dramatic and systematic for the third dimension, suggesting less commonality of structure for the third dimension. Thus, this analysis supports the conclusions drawn from the zyxwvutsrqpo Table 6. Correlations of Dimension Coordinates Between Each Group’s MDS 11-year-olds Ape Group Certainty Adult Monitors Nonmonitors 9-year-olds Monitors Nonmonitors zyxwvutsrqpo zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA II-year-olds .92 .92 1.00 .94 1.00 Monitors Nonmonitors .84 .83 .88 .84 .84 .90 Information Processing II -year-olds Monitors Nonmonitors .91 .82 Monitors Nonmonitors 9- y ear- olds 1.00 .72 1.00 1.00 1.00 .86 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPON 9- y ear- olds Monitors Nonmonitors .78 .83 .75 .89 .83 Dimension 3 II -year-olds Monitors Nonmonitors .39 SO 1.00 .64 1.00 .68 .47 .40 .65 .53 .78 .80 1.00 .74 1.00 1.00 .30 1.00 9- y ear- olds Monitors Nonmonitors Note: rcrit = .41, one-tailed,p < .0.5. Verbs of Knowing 283 previous analysis which suggests a great deal of commonality of structure for the groups along both the certainty and information zyxwvutsrqponmlkjihgfedcba processing dimensions. Combined Data. Given that all analyses suggested that commonality of structure existed between our groups’ similarity judgments for the verbs on at least the first two dimensions, we conducted an INDSCAL analysis to assess whether there were both individual differences and developmental differences in the importance attached to the certainty and information processing dimensions. Our primary hypothesis in this study was that children who monitor their comprehension should be more sensitive to the certainty features of mental verbs compared to their age cohorts. The threedimensional solution was selected because the fit provided by the three-dimensional solution (r = .720, stress .177) was much better than the two-dimensional solution (r = .615, stress = .264). Moreover, the variance accounted for by the three-dimensional solution was more balanced across groups than the two-dimensional solution was. Because the first two dimensions of the INDSCAL showed commonality of structure, Figure 2 presents the organization of mental verbs around the first two dimensions of the three-dimensional INDSCAL. As expected, the first dimension to emerge was the certainty dimension. The ordering of items along this dimension correlated .93 with the certainty, .68 with the inference, and .65 with the information availability scales, all p‘s < .05. A multiple correlation analysis regressing these three attribute ratings together on the position of items along this dimension yielded a multiple R = .98, F (3, 16) = 89.03, accounting for 95.4% of the variance in the organization of verbs along this dimension. The second dimension was the information processing dimension as indicated by the significant correlation of the information processing (.65), perceptual/conceptual (.77), and external/internal (.80) scales, all p’s < .05. Together, a multiple regression of these attributes on the position of verbs along this second dimension yielded R = .84, F (3, 16) = 10.26, p < 05, accounting for 70.3% of the variance in the ordering of verbs along this dimension. Finally, the third dimension seemed to reflect a distinction between verbs according to their visual/spatial-auditory/verbal characteristics (r = .71,p < .05). This dimension seemed to be weighted more for the younger children than the adults. However, because there was no evidence for commonality of structure for this dimension, we place little emphasis on this finding. As predicted, the weight placed on the certainty dimension increased with age and comprehension monitoring status (see Table 7). Although we know of no accepted way of determining significance of the difference along the certainty dimension, we can calculate the binomial p. that the rank 284 P. j. Schwanenflugel, W. V. Fabricius, and C. R. Noyes Output/ Conceptual! Internal zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQ Certain/ Noninferential/ Available In& PerceptuaL! External Figure 2. Certainty dimension (horizontal) and Information-Processing dimension (vertical) from the three-dimensional combined INDSCAL representation of the verb similarity judgments from Experiment 2. (Dimension 3, not depicted, distinguishes visua~ spatia~ from auditory/verbal items.) ordering of the weights of the five groups would both replicate the developmental findings from Study 3 and support our hypothesis that at each age comprehension monitors would weight the certainty dimension more heavily than nonmonitors. The observed ordering meets these conditions, as would the ordering with the ll-year-old nonmonitors and g-year-old monitors switched. The binomial p. that either of these orders would have occurred by chance is ,017. The groups were fairly similar in the relative weights placed on the information-processing dimension. As before, we can also compare relative weightings obtained on the INDSCAL within each group. Examining the weightings in Table 7 in this way, it can be seen that from the 9-year-old nonmonitors to adults, the balance progressively shifts from the ~~2~~~~~~~~~ processing toward the certainty dimension, This also replicates the findings from Experiment 1. 285 Verbs of Knowing Table 7. Group Dimension for Experiment 2 Weights for Three-Dimensional INDSCAL zyxwvutsrqp Analysis Dimension Information Group Adults 11-year-old monitors ll-year-old nonmonitors g-year-old monitors g-year-old nonmonitors Certainty Processing Third .724 642 S67 .490 .421 ,421 ,478 .546 ,532 ,529 ,235 ,320 ,378 ,387 .457 Thus far, we have presented no direct evidence regarding whether children actually understand the certainty aspects of the verbs used in the study other than presenting correlations between adult ratings of certainty and the children’s organization of the cognitive verbs. It is possible that children’s organization of verbs along the certainty dimension resulted either from use of a different feature that happened to be highly related to certainty, or from only an implicit understanding of the different degrees of certainty involved in various mental processes. To obtain more direct evidence that children understood verb certainty, we asked fifteen 9-year-olds and sixteen 11-year-olds to rate a large number of mental verbs on degree of certainty. We, then, compared the average of the children’s ratings for these 17 verbs with the adult certainty ratings. Children’s certainty ratings were remarkably similar to those of adults. Nine-year-olds’ ratings correlated .90 with adults’ ratings. Eleven-year-olds’ ratings correlated 237 with adults’ ratings. Moreover, the children’s certainty ratings were predictive of their age cohorts’ organization along the first dimension of the MDS. Nine-year-olds’ certainty ratings correlated .79 and .71 with the organization along the certainty dimension of the monitors and nonmonitors, respectively. Eleven-year-olds’ ratings correlated 23.5 and .80 with the organization of the monitors and nonmonitors, respectively. Thus, it can be concluded that: (a) children used the feature of degree of certainty to organize mental verbs; (b) their explicit understanding of this feature is remarkably similar to that of adults; and (c) changes in the organization of the cognitive verbs reflect an increase in the emphasis that children place on degree of certainty of mental processes with development, and not changes in understanding of certainty itself. Finally, it is important to discern whether the 9-year-olds possessed some understanding of the moderate frequency items used in the study, that is, those with a frequency 10 to 50 per million in running text zyxwvutsrqponmlkjihgfed (memorizing, observing, recognizing, examining, exploring, estimating, and comparing). As for the previous experiment, we compared 9-year-olds’ MDS placement of 286 P. J. Schwanenflugel, W. V. Fabricius, and C. R. Noyes these verbs with the adult solution. The 9-year-old comprehension monitors’ MDS placed all but zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIH comparing with at least two out of three of the nearest neighbors shown on the adult MDS. Comparing shared only one nearest neighbor. The comprehension nonmonitors’ MDS placed all but estimating and comparing with two out of three of the nearest neighbors on the adult MDS. Estimating shared only one out of three of the adult MDS nearest neighbors, and comparing shared none. Thus, with the exception of comparing, this analysis suggests that the 9-year-olds’ understanding of these moderate frequency items was similar to that of the adults. DISCUSSION These studies show that one of the robust organizational changes to occur in elementary school children’s concepts of mental activities is the increasing importance that children place on the certainty aspects of mental activity. There were both developmental and individual differences in this understanding. At each age, comprehension monitors weighted the certainty aspects of mental activity more than their comprehension nonmonitoring counterparts. This emphasis on the relative certainty of mental processes replicates the findings of Schwanenflugel, Fabricius, Noyes, et al. (1994) for adults. One question that must be addressed is why the relative certainty of mental processes is so important that it serves as a framework around which people organize cognitive verbs. One possibility stems from the idea that certainty and uncertainty represent fundamental cognitive feelings which can be used to guide actions and decisions (Clore & Parrott, 1994; Damasio, 1994). Recent discussions of the role of affect in cognition and reasoning have suggested that affective reactions are vital for producing “smart” cognitive behaviors. Cognitive feelings such as uncertainty may enable people to evaluate their knowledge, understanding, or expectations in order to motivate and modify further action. People tend to use relative certainty of various approaches to decide the course of actions in problem solving (Bechara, Damasio, Damasio, & Anderson, 1994; Bransford, Sherwood, Vye, & Reiser, 1986). Producing uncertainty appears to generate information seeking and directs attention in learning settings (Berlyne, 1960, 1972). Recently, Clore and Parrot (1994) have demonstrated that feelings of uncertainty are also used to monitor comprehension. In that study, adults who had hypnotically-induced, unexplained feelings of uncertainty rated their comprehension of a poem lower than adults who had been told that their uncertainty had come from hypnosis. Clore and Parrott argued that cognitive feelings such as certainty and uncertainty are an “ever-present and well-integrated part of normal metacognitive judgments” (p. 102). Thus, feelings of certainty and uncertainty provide people with vital information Verbs of Knowing 287 for evaluating knowledge states and courses of action and for directing subsequent processing. The increasing importance placed on certainty aspects of cognitive verbs with age may reflect children’s increasing use of feelings of certainty and uncertainty in cognitive activities. When feelings of uncertainty guide processing, people will tend to process more deeply, and to experience those processes as emotionally-laden. This should lead them to see important differences among mental processes on the basis of the degree of uncertainty they entail. Older children are probably more likely to use feelings of uncertainty to guide processing, which would explain the greater weight they give to the certainty dimension. The comprehension monitoring results confirm this presumed developmental connection between feelings, processing, and weight given to the certainty dimension. Comprehension monitoring is a prime example of using feelings of uncertainty to guide processes. Monitors pay more attention to feelings of uncertainty while hearing the stories, attribute feelings of uncertainty to the story, and attempt to check and justify these feelings with reasoning processes such as complex questions and causal statements, which they use to identify the contradictions in the stories. Correspondingly, they place greater weight on the certainty dimension. What is the developmental progression by which children achieve an understanding of the relative certainty of mental processes? Our findings indicate that by 8 years of age, children have a reasonably good understanding of the relative uncertainty of mental processes. However, prior to this time, there is considerable development in this understanding. Consider the verbs zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA know, think, and guess, which anchored the ends and middle of the certainty dimension in our studies. Preschoolers treat these verbs as representing a dichotomy between knowledge and lack of knowledge. For example, Moore, Bryant, & Furrow (1989) showed that 4-year-olds distinguish know from think and guess, but not think from guess. In this dichotomy, know refers to having correct information, whereas think and guess refer to having incorrect and no information, respectively. This dichotomous view of knowledge may represent a fairly general stance that preschoolers have towards the mind. This dichotomous view can be seen in standard early theory of mind tasks-false belief, appearance/reality, and representational change tasks. Four-year-olds know in these tasks that something can be represented two ways (Flavell, 1986), but only in the sense in which one representation is true and the other is false (Flavell, Miller, & Miller, 1993). Young children understand that a false representation results from the absence of crucial information that would lead to correct knowledge. Thus, they can be said to use “true/false” reasoning in regard to knowledge, to believe that things have one true meaning, and to understand that incorrect knowledge comes from absence of information. In contrast, understanding the relative certainty of mental processes 288 P. J. Schwanenflugel, W. V. Fabricius, and C. R. Noyes requires more than a dichotomy of knowledge into true and false. It requires understanding statements such as, “I know x is true,” versus “I think x is true but y and z are also possible,” versus “I guess because I’m totally uncertain and I can’t tell.” It requires conceptual change from a dichotomous distinction between knowledge versus ignorance based on the view that things can have only one meaning, to a continuum based on certainty of knowledge because things can have multiple meanings. Reconceiving knowledge as varying along a continuum of certainty should give the child some insight into why knowledge is not a direct copy of information, and thereby lay part of the foundation for a constructivist theory of mind. What might be the developmental mechanism for changing from a “true/false” view of knowledge to a “degrees of certainty” view? We think that one potentially important mechanism is children’s subjectively experienced feelings of uncertainty. For example, in referential communication tasks (e.g., Beal & Flavell, 1982; Flavell, Speer, Green, & August, 1981; Patterson, Cosgrove, & O’Brien, 1980) which involve responding to and evaluating messages that ambiguously refer to two or more referents, children from ages 4 to 10 will often hesitate, look puzzled, or hover their two hands over equally likely referents. As Clore and Parrott (1994) have shown, adults search for a cause to which to attribute feelings of uncertainty and the same is likely to be true for children. If children have a dichotomous view of knowledge, they could attribute feelings of uncertainty only to error or ignorance. However, in referential communication tasks and in everyday instances of verbal and perceptual ambiguity, both interpretations of the message are equally plausible, so attributions to error or ignorance are impossible. This accounts for the typical error at these ages: Finding a plausible interpretation, children discount their feelings of uncertainty and decide that they are correct. But it is likely that feelings of uncertainty in ambiguous situations persist and demand explanation (see Acredolo & O’Conner, 1991, for a review). Everyday experiences of feelings of uncertainty in response to ambiguity may thus encourage children to construct a new category of “possibly true,” in order to explain their uncertainty, and eventually to construct the continuum of degrees of certainty. In our view, understanding the import of certainty and uncertainty in mental activities represents only a part of developing constructivist theory of mind. A constructivist theory of mind is achieved when children consolidate the following insights: (a) that knowledge can be more or less certain, (b) that feelings of uncertainty are important in evaluating information, (c) that things can have multiple meanings, and (d) perhaps most importantly that those meanings can arise solely from differences in interpretive mental processes that can occur during information acquisition as well as during subsequent processing (Fabricius & Schwanenflugel, 1994). Researchers (e.g., Perner & Davies, 1991; Ruffman, Olson, & Astington, 1991; Sodian, Verbs of Knowing 289 1990) have tended to view children’s success on referential and perceptual ambiguity tasks as indicating they understand that things can have multiple meanings and that the mind interprets information. However, in an ambiguous situation, the information itself is insufficient to draw a conclusive interpretation. In our view, if children recognize representational differences only when sufficient or correct information is absent, then they still have a nonconstructivist theory of mind. We think that increasing weight given to certainty with age also reflects children’s growing understanding that multiple interpretations are common because they result from mental processes and not only from insufficient information. The children and adults also organized mental verbs according to their information zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCB processing characteristics. Organizing mental verbs according to their information-processing aspects also replicates the findings of Schwanenflugel,Fabricius,Noyes, et al. (1994) for adults as well as the distinction made between conceptual and perceptual activities found by Fabricius, et al. (1989). However, our data did not provide clear evidence for the development of these features during the elementary school years, but instead suggested that all groups possessed at least some knowledge of these features. The information-processing distinction may be difficult to distinguish from the simpler one between perceptual activities and conceptual activities, and may require other methods. What would a sophisticated understanding of the information processing characteristics entail that was not identified by our current methods? It would entail, among other things, knowledge of the selective nature of input processes and the interactive nature of mental activities. For example,in our earlier studies (Fabricius et al., 1989;Schwanenflugel Fabricius, & Alexander, 1994), we asked adults and children to rate the similarity of scenarios depicting common, everyday examples of several mental activities including selective attention activities. Using these methods, we were able to identify that children did not understand the selective nature of selective attention (although they did understand the sensory characteristics), and therefore, did not understand that cognitive processes operate as information is acquired. In the current studies, only pay ing attention and noticing related to these selective processes directly. These verbs appeared at the input end with the perceptual verbs hear and see for the children and adults. Thus, although from other work we know that adults understand more than just that these verbs relate to the perceptual characteristics of information processing, the fact that these verbs are also sensory in nature would cause both children and adults to place the verbs at the input end of the information-processing dimension. Thus, it will only be through multiple methods that we come to understand the full nature of children’s versus adults’ understanding of the information-processing characteristics of mental activities. In the study presented here, we have learned that both children and adults find some of these basic information-processing distinctions an important organ- 290 P. J. Schwanenflugel, W. V. Fabricius, and C. R. Noyes izational feature of mental activities. The results of the current study in conjunction with those of our previous ones suggest that some of these information-processing distinctions are basic ones made early by children. Examining the organization of mental verbs in this way has enabled us to expand our understanding of the development of cognitive words in ways not evident in other methods. First, our methods allowed us to discern developmentally the relative importance that various age groups placed on the features of cognitive verbs, which is not evident from previous methods. Second, Hall and his colleagues (Frank & Hall, 1991; Hall et al., 1987; Hall & Nagy, 1986) argue that the cognitive lexicon makes distinctions between perceptual and various conceptual uses (such as memory, understanding, evaluation, metacognition, and planning). However, our studies allow us to assess how children themselves organize the cognitive lexicon. Moreover, similarity ratings allow children and adults to display knowledge of several features of the domain simultaneously such that words like zyxwvutsrqponmlkjihgfedc search can be organized both as highly perceptual and relatively uncertain (which would constitute distinct categories in Hall’s classification). Finally, previous research by Moore, et al. (1989) suggested that 8-year-olds have difficulty understanding the distinction among the three levels of certainty expressed by know, think, and guess, particularly the distinction between think and guess. In our study, placement of these verbs on the individual groups’ MDSs’ showed that even the 8-year-olds placed thinking clearly midway between knowing and guessing in terms of certainty, as did adults. Further, direct ratings of certainty by 9- and ll-year-olds clearly revealed an adultlike distinction among these verbs and other verbs along this attribute. The difference between Moore’s results and ours may be due to the different tasks. Moore et al. (1989) used a communication task in which one speaker said something about a hidden object, for example, “I know it’s in the red box” whereas the other speaker said, “I guess it’s in the blue box.” The child chose where to look based on the relative certainty of the contrasted verbs. The advantage of this task is that it assesses children’s understanding of certainty as a subjectively experienced mental state on the part of the speaker, unlike previous studies (e.g., Abbeduto & Rosenberg, 1985; Johnson & Maratsos, 1977; Olson & Torrance, 1987; Sodian & Wimmer, 1987) in which children made linguistic judgments about the truth value of these verbs or their proper application to informational conditions. However, a potential problem with this task is that think and guess are often used idiomatically (Shatz, Wellman, & Silber, 1983) in expressions such as “I think so,” and “I guess so,” which do not differentiate among levels of certainty. A tendency to interpret the speakers’ statements idiomaticly was made evident by the fact that Moore and Furrow (1991) also found that a substantial proportion of adults tested in this task did not distinguish between think and guess either. Our methods of verb similarity judgments and explicit ratings of certainty 291 Verbs of Knowing also assess understanding of certainty as a mental state, but are likely to have avoided unintended references to idiomatic uses. The methods used in our experiments were designed to uncover robust and salient features of the developing folk theory of knowing. Children certainly have additional knowledge of this domain and, perhaps, additional understandings that also organize this folk theory. In these experiments, we only examined the organization of 25 verbs that children already possessed in their vocabulary. No doubt there would be additional insights gained from examining both the acquisition of new vocabulary for mental activities as well as the application of their vocabulary to an array of contexts. In sum, children in these studies displayed an organization of mental verbs similar to the organization displayed by adults. Specifically, they organized mental verbs according to their certainty and information-processing characteristic. However, children assigned greater importance to the certainty characteristics with age and with comprehension-monitoring status. We have argued that this is reflective of the development of a constructivist theory of mind. Abbeduto, L., & Rosenburg, S. (1985). Children’s knowledge of the presuppositions of know and other cognitive verbs. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONML Journal of Child Language, 12,621- &I. Acredolo, C., & O’Connor, J. (1991). On the difficulty of detecting cognitive uncertainty. Human Development, 34,204- 223. Beal, CR., & FIavell, J.H. (1982). Effect of increasing the salience of message ~biguities on kindergartners’ evaluations of communicative success and message adequacy, Developmental Psy chology , 18,43X% Bechara, A., Damasio, A.R., Damasio, II., & Anderson, S.W. (1994). Insensitivity to future consequences following damage to human prefrontal cortex. Cognition, 50,7- 15. Berlin, El. (1992). Ethnob~ologicaf ~lassi~cat~on: Principles of categorization of plants and animals in traditional societies. Princeton NJ: Princeton University Press. Berlyne, DE. (1960). Conf[icf, arousal, and curiosity . New York: McGraw-Hill. Berlyne, D.E., & Normore, L.F. (1972). Effects of prior uncertainty on incidental free recall. Journal of Experimental Psy chology , 96,434s. Blount, B.G., & Schwanenflugel, PJ. (1993). Cultural bases of folk classificational systems. In J. Altarriba (Ed.), Cognition and czthure. Amsterdam: North Holland. Booth, J.R., & Hall, W.S. (1994). Role of the cognitive internal state lexicon in reading comprehension. Journal of Educational Psy chology , 86,413- 422. Borg, W.R., & Gall, M.D. (1983). Educational research: an introduction. New York: Longman. Bransford, J., Sherwood, R., Vye, N., & Reiser, J. (1986). Teaching thinking and problem solving: Research foundations American PsychoIog~t, 41,1078- 1089. Carroll, J.B., Davies, I?, & Richman, B. (1971). The American Heritage W ord Frequency Book. NY Houghton Mifflin. Clark, A. (1987). From folk psychology to naive psychology. Cognitive Science, 11,139- 154. Clore, G.L., & Parrott, W.G. (1994). Cognitive feelings and metacognitive judgments. European Journal of Social Psy chology , 24,101- 115. Damasio, A.R. (1994). Descarfes error: Emotion, reason, and the human brain NY Putnam. D’Andrade, R. (1987). A folk model of the mind. In D. Hotland & N. Quinn (Eds.), Cultural models in language and thought (pp. 112-148). Cambridge: Cambridge University Press. P. J. Schwanenflugel, W. V. Fabricius, and C. R. Noyes 292 Fabricius, W.F., & Schwanen~ugel, E?J.(1994). The older child’s theory of mind. In A. Demetriou & A. Efklides (Eds), zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJ The structure and development of mind: M odels, methods, and applications. Amsterdam, Elsevier. Fabricius, W.F., Schwanenflugel, PJ., Kyllonen, PC., Barclay, C.R., & Denton, SM. (1989). Developing theories of the mind: Children’s and adults’ concepts of mental activities. zyxwvutsrq Child Development, 60,127&- 1290. Flavell, J.H. (1986). The development of children’s knowledge about the appe~ance-reality distinction. American Psy chologist, 41,418- 425. Flavell, J.H., Speer, J.B., Green,F.L., & August, D.L. (1981). The development of comprehension monitoring and knowledge about communication. M onographs of the Society for Research in Child Development, 46 (Serial No. 192), l-6.5. Flavell, J.H., Miller, RH., & Miller, S.A. (1993). Cognitive developnzent. Englewood Cliffs, NJ: Prentice-Hall. Frank, R.E., & Hall, W.S. (1991). Polysemy and the acquisition of the cognitive internal state lexicon. Journal of Psy cholinguistic Research, 20,283- 3&I. Frye, D., & Moore, C (1991). Children’s theory of mind: M ental states and social understanding. Hi&dale, NJ: Eribaum. Hall, W.S., & Nagy, W.E. (1986). Theoretical issues in the investigation of words of internal report. In I. Gopnik & M. Gopnik (Eds), From models to moduEes:Studies in cognitive science from the M cGill workshops (pp. 25-65). Norwood, NJ: Ablex. Hall, W.S., Scholnick, E.K., & Hughes, A.T. (1987). Contextual constraints on usage of cognitive words. Journal of Psy cholinguistic Research, l&289- 310. Johnson, C.N., & Maratsos, M.P. (1977). Early comprehension of mental verbs: Think and know. Child Development, 48,1743- 1747. Keesing, R.M. (1987). Models, “folk” and “cultural”: Paradigms regained? In D. Holland & N. Quinn (Eds.), Cultural models in language and thought (pp. 369-393). Cambridge,: Cambridge University Press. Kiparsky, P., & Kiparsky, C. (1970). Fact. In M. Bierwisch & K. Heidolph (Eds), Progress in ii?zgu~stjcs(pp. 143-173). The Hague: Mouton. Kruskal, J. B., & Wish, M. (1978). M ultidimensional scaling. Sage University Paper Series on Quantitative Applications in the Social Sciences (Series No. 07-011). Beverly Hills, CA: Sage. Markman, E.M. (1981). Realizing that you don’t understand: Elementary school children’s awareness of inconsistencies. Child Development, SO, 643- 455. Miller, K. (1987). Geometric methods in developmental research. In J. Bisanz, C.J. Brainerd, & R. Kail (Eda), Formal methods in developmental psy chology : Progress in development research (pp. 216-262). NY Springer-Verlag. Moore, C, & Furrow, D. (1991). The development of language of beliefi The expression of relative certainty. In D. Frye & C Moore (Eds), Children’s theories of mind: M ental states and social ~derstanding (pp. 173-193). Hillsdale, NJ: Erlbaum. Moore, C., Bryant, D., & Furrow, D. (1989). Mental terms and the development of certainty. Child Development, 60,167- 171. Murphy, G.L., & Medin, D.L. (1985). The role of theories in conceptual coherence. Psy chologi- cal Review, 92,289- 316. Oison, D.R., & Torrance, N. (1987). Language, literacy, and mental states. Discourse Processes, 10, f57-167. Patterson, C.J., Cosgrove, J.M., & O’Brien, R.G. (1980). Nonverbal indicants of comprehension and noncomprehension in children. Developmental Psy chology , 16,38- 48. Perner, J. (1991). Understanding the representational mind. Cambridge, MA: MIT Press. Pemer, J., & Davies, G. (1991). Understanding the mind as an active information processor: Do young children have a “copy theory of mind”? Cognition, 39,.5169. SAS Supplemental Library User’s Guide (1980). Carey, NC: SAS Institute Inc. 293 Verbs of Knowing Ruffman, T., Olson, D.R., & Astington, J.W. (1991). Children’s understanding of visual ambiguity. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA British Journal of Developmental Psy chology , 9,89- 102. Schwanenflugel, P.J., Fabricius, W.F., & Alexander, J.M. (1994). Developing theories of mind: Understanding concepts and relations among mental activities. Child Development, 65, 1546- 1563. Schwanenflugel, P.J. Fabricius, organization of mental W.F., Noyes, C.R., Bigler, K., & Alexander, J.M. (1994). The verbs and folk theories of knowing. Journal of M emory and Language, 33,376- 395. Shatz, M., Wellman, H.M., & Silber, S. (1983). The acquisition of mental verbs: A systematic investigation of the first reference to mental state. Cognition, 14,301-321. Shepard, R.N. (1972). A taxonomy of some principal types of data and of multidimensional methods for their analysis. In R.N. Shepard, A.K. Romney, & S.N. Nerlove (Eds), M ultidimensional scaling: Theory and application in the behavioral sciences (Vol. 1, pp. 2347). New York: Seminar Press. Shoben, E.J., & Ross, B.H. (1987). Structure and process in cognitive psychology using multidimensional scaling and related techniques. In R. Ronning, J. Glover, J.C. Conoley, & J.C. Witt (Eds.), The influence ofcognitivepsy chology on testing (pp. 229-266). Hillsdale, NJ: Erlbaum. Sodian, B. (1990). Understanding verbal communication: Children’s ability to deliberately manipulate ambiguity in referential messages Cognitive Development, 5,209- 222. Sodian, B., & Wimmer, H. (1987). Children’s understanding of inference as a source of knowledge. Child Development, SZ424- 433. Wellman, H.M. (1990). The child’s theory of mind. Boston: MIT Press. Wing, C.S., & Scholnick, E.K. (1986). Understanding the language of reasoning: Cognitive, linguistic, and developmental influences. Journal of Psy cholinguistic Research, IS, 383401. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED APPENDIX Examples Experiment provided to Children during the Instruction Phase 1 Verb Example Checking Going over your homework to see if you have made a mistake. Making up your mind about what to watch on TV. Finding a shoe you though you’d lost at the bottom of the toy box. Telling your Mom why you got a C in math. Trying to figure out what your teacher is going to ask you next. Listening to some music. Making up a new game. Raising your hand to give the answer to a question your teacher asks. Going over the multiplication table in your head. Picking out a word that is misspelled in a paragraph. Deciding Discovering Explaining Guessing Hearing Inventing Knowing Memorizing Noticing 294 Paying Attention Thinking Understanding P. j. Schwanenflugel, W. V. Fabricius, and C. R. Noyes Listening to the announcements in a noisy school cafeteria. Wondering why there are so many stars up in the sky at night. Feeling like you know how to do an assignment after the teacher explains it. Experiment 2 Comparing Estimating Examining Exploring Guessing Knowing Learning Memorizing Observing Questioning Reasoning Recognizing Remembering Searching Seeing Thinking Understanding Finding two people who are the same height for a science project. Trying to figure out how tall your friend is without using a yardstick. Looking at a stamp through a magnifying glass in science. Wandering through a maze or looking for clues to a puzzle. Your friend coming up behind you and saying “Guess who?” Answering a question correctly. Listening to your teacher explain how to do something until you know how to do it. Practicing your friend’s phone number until you don’t have to look it up in the phone book. Looking at the monkeys in their cages at the zoo. Asking a person what their name is. Coming up with an explanation for why something happened. Seeing your kindergarten teacher at the mall. Bringing your best friend a present for her birthday. Trying to find your dog that ran away. Watching a movie. Coming up with an idea for a story. Following the instructions on a box.