Margaret Shavlik Thesis

Running head: FURNISHING THE MIND 1

Furnishing the Mind: Understanding How Children Learn Adjectives

Margaret Shavlik

Faculty Advisor: Sandra Waxman

Second Reader: Steven Franconeri

Northwestern University

Department of Psychology – Honors Thesis

FURNISHING THE MIND 2

Abstract

An intuition-based theory of adjective learning is that it is a perceptually-based process.

However, it seems to be the case that children are not forming mere associations between a word

and a perceptual experience, but are in fact incorporating reference from their holistic knowledge

of the world. The current study put this to the test. We showed children books of ambiguous

shapes (e.g., a shape that could be interpreted either as spilled paint or a painting of a dragonfly),

introducing the books as full of “pictures of things” or “blobs of stuff.” On each page, we

described a target shape using a novel adjective (e.g., “blickish”), and as we showed two test

shapes (one of which matched the first in color) we asked the child if they could find “another

one that is blickish.” We found that this difference mattered: children who thought the shapes

were “blobs of stuff” were much more accurate at indicating a color match than those who

thought that the identical shapes were “pictures of things.” Put in context, the “blue” property a

child extracts from a dragonfly seems to be different—and less flexible—than the “blue”

property she extracts from a dragonfly-shaped puddle. Conceptual information, not just

perceptual information, seems to play a role in how children understand adjectives.

Acknowledgments My sincere thanks to my advising professor, Sandra Waxman, and graduate student advisor,

Alexander Latourette. I would also like to thank Brooke Sprague, Jennie Woodring, and

everyone at the Project on Child Development for their help and support.


Furnishing the Mind: Understanding How Children Learn Adjectives

Literature Review

Childhood language skills are a strong predictor of overall success in school and work

(Biemiller, 2006; Biemiller, 2001; Song et al. 2015) and understanding what facilitates early

language learning could offer tools for addressing the “vocabulary gap” that exists between those

with high and low childhood vocabularies (Biemiller & Slonim, 2001; Chall, Jacobs, & Baldwin,

2009). Communication is the building block upon which much other learning relies, and as such,

understanding the mechanisms by which language learning occurs is vital.

Infants gain the skill to distinguish between different grammatical categories (nouns,

verbs, etc.) not long after speaking their first words – and long before they start using them in

sentences (Waxman & Booth, 2001). This ability to distinguish parts of speech is instrumental in

helping children learn to match the words they hear with the appropriate referents in the world

around them. Children's amazing word-learning capacity then serves as an important cornerstone

in their cognitive and linguistic development (Waxman & Markow, 1995; Biemiller, 2006).

Arguably, the hallmark social, conceptual, and linguistic abilities of the human mind share a

common lineage: word learning. Early in this process, children learn words for objects (e.g., dog)

and properties (e.g., white; Waxman & Booth, 2001).

An intuition-based theory of adjective acquisition is that it is a perceptually-based process.

A child hears a word, sees the property being described, and thus, is able to use the adjective to

describe other instances of that property. Philosopher John Locke articulates this approach,

famously describing the human mind as “white paper, void of all characters, without any ideas.”

According to Locke, the immense task of “furnishing” the mind relied simply on experience.


Thus, Locke argued:

The same colour being observed to-day in chalk or snow, which the mind yesterday received from milk, it considers that appearance alone…and having given it the name whiteness, it by that sound signifies the same quality wheresoever to be imagined or met with. (Locke, 1690/1975)

One common instance of this furnishing, like in Locke’s example, involves the process by

which a child learns the names for colors. If Locke were right, then after hearing “Look at this

brown dog,” it should be easy for a young child to identify a house that is brown. However, this

process is in actuality much more complex for young language learners. First, they need to

identify which feature of the dog is described by “brown” (color? texture? other?). Next, children

have to learn how “brown” can be applied to other objects from the same category (e.g., from

one dog to another), and then, objects from new categories (e.g., from dogs to houses). This step

is surprisingly difficult for young learners (Sandhofer & Smith, 1999; Klibanoff & Waxman,

2000), which implies that children are not just assigning a perceptual quality a name. Instead, it

is likely that children are incorporating conceptual information into their understanding of a new

adjective’s referent. Just as is the case when learning names for object-nouns (see Waxman &

Gelman, 2009 for overview), children are not forming mere associations between a word and a

perceptual experience, but are in fact incorporating reference from their holistic knowledge of

the world.

Research on the role of intention and analogy in children’s naming of pictorial

representations supports this account as well (e.g., Gelman & Bloom, 2000; Christie, Gentner,

Vosniadou, & Kaiser, 2007). Sameness of shape has been found to be neither necessary nor

sufficient for two representations to share a name (for example, while an egg and a football are

quite similar in shape, most people would agree that they should not share an object name).


Children make this distinction as well when viewing accidentally versus intentionally created

drawings: they much more likely to name the resulting shape if it was intentionally created than

if it were created on accident (Bloom & Markson, 1998).

To illustrate the role of conceptual information in adjective extension, Klibanoff &

Waxman (2000) taught children novel adjectives (e.g., “blickish”; see appendix) to describe

familiar objects. They found that adjective learning is not simply matching one hue to another.

Instead, children’s interpretation of property terms (adjectives) varies with the kind of object

being modified by the term (e.g., the “softness” of a kitten differs from the “softness” of a scarf).

If adjective learning was solely a perception-based process, children should find it equally easy

to extend a property name (e.g., “blickish”) for a quality (such as “being-bumpy’) from an apple

to another apple, or from an apple to a boat. It would not matter if two “blickish” objects were of

different or the same shape, as long as the salient perceptual feature (texture) was present in at

least one of the test items. However, this was not the case, as Klibanoff and Waxman found that

these property words were much easier to map from the trial object onto a test object of the same

basic-kind. This suggests that perceptual features may be initially constrained by item category

when children are first learning the meaning of the property.

In subsequent pilot testing, Waxman (2002) took this logic a step further. Would this

shape-bias still exist if children saw ambiguous shapes, instead of known objects? And could this

bias be manipulated solely based on conceptual information (how the stimuli were introduced)

while keeping all perceptual information constant? She reasoned that children would be more

successful at extending a novel adjective if it has been introduced as a property of a substance

(“a blickish blob of stuff”) than as a kind of object (“a blickish picture of a thing”), because all

blobs would fit under the same overarching category whereas each kind of object has its own


category membership and rules. She developed ambiguous stimuli which varied in color and

texture (see Appendix 0.2.1 for an example), introducing them to some children as “pictures of

things” and to others as “blobs of stuff.” This change in wording had a striking effect: children

successfully extended a novel adjective from one “blob of stuff” onto any other, but failed to do

so when the same ambiguous image was described as a “kind of thing.” This pilot is the

foundation for the current study. We updated (and validated) the stimuli through further pilot

testing and increased sample size. We predicted that her findings would hold: that children

would only be able to extend new adjectives onto objects of the same basic-kind, but to any kind

of non-object.

In the current experiment, we created books of stimuli that were either introduced as “a

book full of lots of pictures of things” or “a book full of lots of blobs of stuff.” We introduced a

target stimulus using a novel adjective (e.g., “blickish,” “zavish”), and asked the child if she or

he could find “another one that is “blickish,” showing two more stimuli (that matched each

others’ shape, but not that of the target) that varied along the parameter of color. We chose to

isolate color (instead of using both color and texture, as was the case in Waxman, 2002) because

it was easier to control and showed the effect of the conceptual manipulation more clearly in the

original study. In either condition, the stimuli were identical (the same perceptual information).

What varied was how the books were introduced (the conceptual information).

Predictions

We predicted that there would be a significant difference between the BLOBS condition

and the other conditions. This would support our hypothesis that there is an overall benefit from

being in the BLOBS condition as opposed to the other two.


Based on the findings of the previous study (Waxman 2002), we predicted no significant

difference between the OBJECTS and the CONTROL conditions. In absence of an explicit

contradiction, children may assume that stimuli are “objects” (i.e., exhibit an object-bias).

While we did not make an explicit prediction in this vein, an interesting question is

whether there would be a significant difference between the CONTROL condition and the other

conditions combined (i.e., “adjectives condition”)? A difference would support existing research

on effects of labeling as a bootstrapping tool. A lack of a difference would suggest that an

adjective-label does not provide a significant advantage in this task.

1. Piloting

Method

Participants

Participants were 30 typically developing monolingual children from the Chicagoland

area. There were 12 boys and 18 girls. Participants were recruited via the lab’s database of

interested families. Most participants were from the Evanston area, of higher socioeconomic

status, and white.

In Study 1a, 18 two and three year olds (25.6 – 37.8 months, mean age 32.0 months,

SD=4.3) participated. Six children from the original subject pool (24 children) were excluded for

never answering the questions (3), failing to complete the task due to distraction (2), and

experimenter error (1). In Study 1b, participants were 12 two and three year olds (28.6 – 36.2

months, mean age 32.2 months). One child was excluded from the original subject pool (14

children) for never answering the questions, and 1 was excluded due to experimenter error.


Measures

Study 1a: Ambiguous shapes that were deemed “blobs” by at least 50% of children were used

in Study 1b. Thirteen ambiguous shapes passed the 50% criterion. One was excluded for being

borderline and looking too similar to another shape.

Study 1b: Ambiguous shapes that were named by at least 50% of children were used in the

books for Experiment 1. All 13 blobs were named as specific objects by at least 55% of the

children (see appendix for names given).

Procedure

Piloting took place at the Project on Child Development at Northwestern University.

Children sat with a female experimenter and their parent in a private room. The task took

approximately 10 minutes.

In study 1a, children were shown a deck of “cards” with silhouettes on them. Half of the

silhouettes were those of clear objects (e.g., a dog), while the other half were ambiguous shapes

created by the experimenters (e.g., a shape that could be a blob, or could be a mitten; see

Appendix 1.1). Children were shown six example cards – three “blobs of stuff” and three

“pictures of things” and told that they were going to play a sorting game. Children were then

asked whether each card had “a blob of stuff” or “a picture of a thing” on it (see script in

Appendix 1.1).

In study 1b, children were shown silhouettes again, half of which were 13 objects from

the set in study 1a (e.g., a dog), while the other half were the 13 ambiguous shapes that met

inclusion criterion from study 1a. Children were told that all of the cards had “pictures of things

on them.” The experimenter explained that for some of the pictures, she knew what was on them,


but for some, she couldn’t tell what the picture was and needed help. Children were presented

one card at a time, and asked “What do you think this is a picture of?” (see script in Appendix

1.2).

Results

By the end of study 1a, thirteen of our ambiguous shapes had been sorted as “blobs” by at

least 50% of participating children. Because two of the shapes were perceptually similar (our

snail and whale) we eliminated one (the snail, which had a lower blob-rating). By the end of

study 1b, all 12 of the remaining shapes had been named as specific things (by new children) at

least 55% of the time.

Discussion

The above results suggest that our shapes truly were ambiguous: contingent upon the

framing of the task, the same shapes could be sorted as either “blobs of stuff” or “pictures of

things.” Thus, we found out stimuli sufficient to use in the experiment proper.

2. Experimental Manipulation

Method

Participants

Participants were 51 typically developing two- to three-year-olds (29.1 – 42.3 months,

mean age 36.0 months, SD=3.7) from the Chicagoland area. Most participants were from the

Evanston area, of higher socioeconomic status, and white. Three from the original subject pool

(53) were excluded for never answering the questions (1), or failing to complete the task due to

distraction (2). Participants were recruited via the lab’s database of interested families and via


letters sent home with parents at the preschool. There were 30 boys and 21 girls. Twelve

children had exposure to a second language at home (defined as >25% non-English heard).

Measures

Performance was measured in terms of test-object choice. Each of the 12 trials were

forced choice responses between two test items. For each participant, “percent-correct choice”

variable was created by tallying their correct responses and dividing by the total number of

responses they gave (which was almost always 12). We submitted the data to a one-way analysis

of variance (ANOVA) with condition (3) as a between subjects condition and percent-correct

choice as our dependent variable.

Procedure

The current experiment followed the logic and procedures of Waxman (2002). The

experiment took place at the Project on Child Development at Northwestern University or in a

private room in a local preschool. In both locations, participants were with the same female

experimenter. At the Project on Child Development, but not at the preschool, the child was

accompanied by his or her parent as well. The task took approximately 10 minutes. Children

looked through a book with the experimenter. Inside the book were the “blob” stimuli validated

during the pilot part of this study (Study 1a and 1b). There were two different versions of the

book, counterbalanced and showing “across basic-kind” triads (test shapes differed from the

target, e.g., two keys and a baseball cap; see appendix for full details). Twenty-six children were

shown Book 3, and 25 were shown Book 4.

In the OBJECTS condition, the experimenter explained to the child that she had a book

“full of pictures of things” and showed an example thing (a dog) and explained that there would


only be pictures of things in her book, and no blobs of stuff (showing an example blob). She then

opened to a page in the book, and pointed to the target shape, saying “Look at this picture! This

is a BLICKISH one. Can you find another one that is BLICKISH?” and then showed the two test

objects. Test objects were on a half-sheet of paper that was lifted up so that the whole triad of

shapes could be seen at once (see photo in Appendix 2.1). The test objects were different shapes

than the target shape, but the same shape as each other. The two shapes only differed in color;

one was the same color as the target (thus, embodying the BLICKISH quality) while the other was

a different color. This continued for 12 triads in each book, with a different novel adjective for

each triad. Participants were randomly assigned to one of the two versions of the book.

In the BLOBS condition, the experimenter explained to the child that she had a book “full

of “blobs of stuff” (and showed an example blob). She explained that there would only be blobs

of stuff in her book, and no pictures of things (showing the dog silhouette). She then opened to a

page in the book, and pointed to the target shape saying “Look at this blob! This is a BLICKISH

one. Can you find another one that is BLICKISH?” and then showed the two test objects. The test

objects were different shapes than the standard, but the same shape as each other. The two shapes

only differed in color; one was the same color as the target (thus, embodying the “blickish”

quality) while the other was a different color. This continued for 12 triads in each book, with a

different novel adjective for each triad. Participants were randomly assigned to one of the two

versions of the book.

In the CONTROL condition, the experimenter introduced her book as “full of pages,”

showing the example dog (“some pages are like this one”) and blob (“and some pages are like

this one”). She then opened to a page in the book, and pointed to the target shape saying “Look

at this page! Look at this one here. Can you find another one like it?” and then showed the two


test objects. This continued for 12 triads in each book. Participants were randomly assigned to

one of the two versions of the book.

Results

As expected, only the condition variable significantly predicted performance. One-way

analyses of variance indicated no effect of stimulus order, second-language exposure, or gender

on choice accuracy (all ps >.25). Additionally, a correlation analysis indicated no effect of age

(r=-.06, p=.69). A regression analysis considering all four of these factors also confirmed

condition as the only significant predictor (ß=.30, p=.047).

For the effect of condition, we hypothesized that children in the BLOBS condition would

be more likely to select the property-match than would those in either the OBJECTS or CONTROL

conditions, and that there would be no difference in the latter two conditions. To test for this

effect, we used a one-way analysis of variance (ANOVA) on participants’ choice accuracy

(likelihood of selecting the property-matched test image). The assumption of homogeneity of

variance was not met, as indicated by Levene’s test (F(2,48)=5.35 p=.008). Thus, we obtained

the Welch’s adjusted F ratio, which indicated a reliable effect of condition, F(2, 30.52)=3.811,

p=.033. Choice accuracy was near ceiling for the BLOBS condition (M=.90, SD=.17) while the

CONTROL (M=.71, SD=.25) and OBJECTS (M=.76, SD=.29) conditions showed somewhat lower

accuracy. Planned orthogonal contrasts demonstrated that, indeed, the BLOBS condition

significantly differed from the other two conditions (t(44.66)=2.65, p=.011, d=.74) while the

OBJECTS and CONTROL conditions did not significantly differ from each other (t(31.65)= -.68,

p=.50, d=.18). These findings support our prediction that perceiving a given shape as a blob,

without an object identity, facilitates adjective learning (see Figure 1).


Individual planned comparisons also supported this conclusion, with the BLOBS condition

outperforming both the OBJECTS (t(26.67)=1.67, p=.11, d=.59) and CONTROL conditions,

(t(28.36)=2.63, p=.014, d=.89). Additional nonparametric (chi-squared) comparisons were

calculated in order to determine whether each condition differed from a chance-distribution of

scores landing either above or below 50% accuracy (the accuracy rate yielded by guessing). Only

performance in the BLOBS condition significantly differed from such a chance distribution

(χ2=8.088, P=.004, other ps>.23). Finally, a chi-square analysis comparing the OBJECTS and

BLOBS distributions on the same measure revealed a marginal difference between the OBJECTS

and BLOBS condition (χ2=3.238, p=.072), providing converging support for the marginal

difference in group means in overall accuracy.

Taken together, these results suggest that encouraging children to view the shapes as

BLOBS, not objects, substantially increases the likelihood that they will extend novel adjectives to

shapes that differ enough to be considered, by children in the OBJECTS condition, as coming from

different basic level kinds.

---------------------------------------------------------------------- Insert Figure 1 Here

----------------------------------------------------------------------


Figure 1 | Performance differences between experimental groups. Toddlers in the BLOBS condition successfully identified the property match more frequently than those in the CONTROL and OBJECTS conditions.

Discussion

In this experiment, we documented an effect of conceptual knowledge on early word

learning. We had predicted that the BLOBS condition would perform better than the other

conditions, which was generally confirmed by both parametric and nonparametric analyses.

These findings suggest that when trying to extend a given shape’s properties to others differing

in shape, there is a benefit to conceiving of that shape as a blob. When shapes are described as

blobs, they do not fall victim to an object-bias, and children are better able to identify

commonalities across basic kinds. When the same shapes are described as objects, children

struggle to identify these commonalities. Furthermore, we had predicted there to be no difference

between the OBJECTS and the CONTROL conditions. Our findings supported this notion,

suggesting that the default assumption for children may be that new stimuli are “objects” (i.e., an

object-bias).


Interestingly, many more children in the THINGS and CONTROL conditions performed

above chance than did so in the original 2002 study. Indeed, this difference represents the most

substantial difference between the original and present findings. There are multiple possibilities

for this elevated overall performance. First and foremost, our stimuli were different than in the

2002 study. In the original study, the shapes varied in either color or texture on each trial,

whereas ours always varied in color only. This may have made the meaning of the adjectives

more obvious to the children in the current study: children may have learned that color was

always the property being mapped. Furthermore, due to the inclusion of texture, children were

encouraged to touch the shapes in the 2002 study, which may have been distracting. Lastly, our

shapes were computer generated and thus more standardized than those in the 2002 study, which

were cut out by hand and thus subject to more variation between each pair of “matching” shapes.

This may have increased the salience of the color dimension, as it was truly the sole

distinguishing feature.

This heightened performance in the present study may also be due to environmental

differences. In the 2002 study, all children were tested in a preschool. In our study, nearly half

were tested at the preschool and half at our research laboratory. As more variables can be

controlled in a lab, and these children were in the comforting presence of their parent, this may

have led to an advantage for these children. Other potential factors include the relative

engagement of children with the 2002 and 2016 experimenters and possible changes in

children’s home experiences in the intervening 14 years (e.g., increased emphasis on color

learning). However, it is important to note that all conditions experienced a similar increase in

accuracy across experiments, resulting in a consistent advantage for the BLOBS condition.

Limitations and Next Steps


There remains a gap in this line of reasoning: how do these “blobs of stuff” or “pictures of

things” arise in children’s everyday experience? (Children are rarely explicitly told “this is an

object” or “this is just a substance”). Without this crucial step, it is difficult to see the importance

of these findings for everyday language learning. Future studies will examine the conditions

under which children interpret images as objects (“kinds of things”) versus non-objects (“blobs

of stuff”). Rather than explicitly telling children that the picture being shown is an object or a

“blob of stuff,” the researcher could show children how it is created—a much more naturalistic

source of information about a shape’s identity

For instance, research on “artifact intention” (the intended use of a created entity) has

indicated that young children have different expectations about intentional versus accidental

events. For example, children are more likely to name a shape if it is created intentionally, as

opposed to accidentally (Gelman & Ebeling, 1998; Gelman & Bloom, 2000). Thus, one could

test whether 3-year-olds can use the intentions of a creator to identify whether an image is an

object (e.g., an ink blot that was purposely created) or a blob (e.g., the same ink blot, created as a

result of an accidental spill). Extending the findings of our current study, I would predict that this

difference will influence infants’ adjective extension in the same manner. In other words, if

children believe that the stimulus was created on purpose (and thus, was an object), they should

restrict their extension of adjectives to other objects of the same kind. However, when the

stimulus is created by accident (e.g., when someone accidentally spills the substance), they

should extend adjectives across many entities (i.e., all shapes that share the same property, such

as color).

Conclusion


Our findings suggest that there is, in fact, an effect of our conceptual manipulation on

adjective extension. Thus, contrary to John Locke’s theory, when learning adjectives, children

are not merely relying on perceptual information. Indeed, it seems that our ability to identify the

properties of a shape depends on the way in which we conceptualize it. The “blue” property we

extract from a dragonfly is different—and less flexible—than the “blue” property we extract

from a dragonfly-shaped puddle. Even if the perceptual information is one and the same. Broadly

speaking, then, conceptual information plays a key role in early word-learning.


References

Biemiller, A. (2001). Teaching vocabulary. American Educator, 25(1), 24-28.

Biemiller, A. (2006). Vocabulary development and instruction: A prerequisite for school

learning. Handbook of early literacy research, 2, 41-51.

Biemiller, A., & Slonim, N. (2001). Estimating root word vocabulary growth in normative and

advantaged populations: Evidence for a common sequence of vocabulary acquisition.

Journal of Educational Psychology, 93(3), 498.

Bloom, P., & Markson, L. (1998). Intention and analogy in children's naming of pictorial

representations. Psychological Science, 9(3), 200-204.

Booth, A. E., Waxman, S. R., & Huang, Y. T. (2005). Conceptual information permeates word

learning in infancy. Developmental psychology, 41(3), 491.

Chall, J. S., Jacobs, V. A., Baldwin, L. E., & Chall, J. S. (2009). The reading crisis: Why poor

children fall behind. Harvard University Press.

Christie, S., Gentner, D., Vosniadou, S., & Kayser, D. (2007). Relational similarity in identity

relation: The role of language. In Proceedings of the Second European Cognitive Science

Conference, ed. S. Vosniadou & D. Kayser. Routledge.[DG].

Gelman, S. A., & Bloom, P. (2000). Young children are sensitive to how an object was created

when deciding what to name it. Cognition, 76(2), 91-103.

Gelman, S. A., & Ebeling, K. S. (1998). Shape and representational status in children's early

naming. Cognition, 66(2), B35-B47.


Klibanoff, R. S., & Waxman, S. R. (2000). Basic Level Object Categories Support the

Acquisition of Novel Adjectives: Evidence from Preschool‐Aged Children. Child

development, 71(3), 649-659.

Locke, J. (1975). An essay concerning human understanding. Oxford, England: Clarendon

Press. (Original work published 1690)

Sandhofer, C. M., & Smith, L. B. (1999). Learning color words involves learning a system of

mappings. Developmental Psychology, 35(3), 668.

Song, S., Su, M., Kang, C., Liu, H., Zhang, Y., McBride‐Chang, C., ... & Shu, H. (2015).

Tracing children's vocabulary development from preschool through the school‐age years:

an 8‐year longitudinal study. Developmental science, 18(1), 119-131.

Waxman, S. R. (2002). Not by perception alone: Conceptual and semantic factors underlying

children’s extension of novel adjectives. In B. Skarabela, S. Fish, & A. H.-J. Do (Eds.),

Proceedings of the 26th Annual Boston University Conference on Language Development

(pp. 746-757). Somerville, MA: Cascadilla Press.

Waxman, S. R., & Booth, A. E. (2001). Seeing pink elephants: Fourteen-month-olds'

interpretations of novel nouns and adjectives. Cognitive psychology, 43(3), 217-242.

Waxman, S. R., & Gelman, S. A. (2009). Early word-learning entails reference, not merely

associations. Trends in cognitive sciences, 13(6), 258-263.

Waxman S. R., & Markow D. B. (1995). Words as invitations to form categories: Evidence from

12-to 13-month-old infants. Cognitive psychology, 29(3), 257-302.


APPENDIX 0 : LITERATURE REVIEW

0.1 Logic of Klibanoff & Waxman (2000)

(Note: these diagrams were created by me to explain the logic of this study, and are not actual stimuli used in the study itself)


0.2 Logic of Waxman (2002) (Note: these diagrams were created by me to explain the logic of this study, and are not actual stimuli used in the study itself)

0.2.1 Examples of actual stimuli used in Waxman (2002).

From left to right: a bird, a tree, a snake.


APPENDIX 1: PILOTING

1.0 Stimulus Validation Scripts 1.1 Blob/Thing Sorting Hey (NAME), look! This book is full of pictures of things (used Tana Hoban’s Of Colors and Things book). Like this one (choose an example in the book). Is that a picture of a thing? Yeah! And what about this one, is this a picture of a thing? Yeah it is! These are all pictures of things. Okay great. So, I have a bunch of different cards. On some of them, they have pictures of things, like this one! (show dog). That’s a picture of a dog, right? And on some of my cards, there aren’t pictures of things. They are just blobs of stuff. Like this one! (show blob). That’s just a blob of stuff, right? Work through 4 more guided examples (2 things, 2 blobs). You’re good at this! Okay, let’s look at more of my cards! I need you to help me decide if each card has a picture of a thing on it, or just a blob of stuff. Are you ready? Great! Do you think this one is a blob of stuff? What about this one? (etc.) Great! Thank you for all of your help! 1.2 Thing Naming Okay (NAME), I have to tell you something. I have a problem. Do you think you can help me?

Great! Thank you! So, I have a problem. I have all of these cards. See? Well, all of these cards have pictures on them. Right?

Well, some of them I know what the picture is of – like this one. That’s a (show dog) right? Good job.

But here’s my problem. Some of the cards, I can’t tell what the picture is! Like this one…. can you help me? What do you think this is a picture of?

Thank you! You are such a good helper. I’m going to write that down. Do you think you could help me with some more?

[REPEAT] Great! Thank you for all of your help!


1.1 Ambiguous Shapes

“key”

“mitten” “whale”

“spoon”

“hammer”

“flashlight”

“baseball hat”

“ice cream cone”

“car”

“flag”

“dragonfly”

“key”


1.2 “Thing” Names How children named the shapes when presented as “pictures of things.” We made sure that shapes that were given the same name were not put in a triad together for the experiment.


APPENDIX 2: Experimental Manipulation 2.0 Experiment Script Pictures of Things: Introduction:

Hey [NAME]! Look, I have a book. Wanna see? Okay but first, let me show you what's gonna be inside my book! (*use two cards - blob and dog). My book is full of pictures of things. Like this one (point at dog). This is a picture of a dog! Right? My book DOESN'T have any blobs of stuff like this one, nope (point to blob, shake head). My book ONLY has pictures of things like this one (point at dog again, smile and nod). Wanna look at my book now?

Task: Look at this picture! (point at target). This is a [BLICKISH] one. Can you find another one that is [BLICKISH]? (lift up test items page) Thank you for helping! (etc)

No Word: Introduction:

Hey [NAME]! Look, I have a book. Wanna see? Okay but first, let me show you what's gonna be inside my book! (*use two cards - blob and dog). My book is full of pages. Like this one (point at dog - switch between using blob first vs thing first). See? My book has some pages like this one (point at blob, nod). And my book has some pages like this one (point at dog again, nod). Wanna look at my book now?

Task: Look at this page! Look at this one here. (point at target) Can you find another one like it? (lift up test items page). Thank you for helping! (etc)

Blobs of Stuff: Introduction:

Hey [NAME]! Look, I have a book. Wanna see? Okay but first, let me show you what's gonna be inside my book! (*use two cards - blob and dog). My book is full of blobs of stuff. Like this one (point at blob). This is just a blob of stuff. Right? My book DOESN'T have any pictures of things like this one, nope (point at dog, shake head). My book ONLY has blobs of stuff like this one (point at blob again, nod). Wanna look at my book now?

Task: Look at this blob! This is a [BLICKISH] one. Can you find another one that is [BLICKISH]? Thank you for helping! (etc)


2.0.1 Adjectives Used

From Klibanoff & Waxman, 2000

Adapted from Yoshida & Hanania, 2013

Adapted from Mintz & Gleitman, 2002

Blickish (KEY) Toop(ish) (DRAGONFLY) rup(ish) (MITTEN) dakish (HAMMER) Vap(ish) (BASEBALL HAT) drin(ish) (FLASHLIGHT) zavish (ICE CREAM CONE) Stoof(ish) (FLAG) prall(ish) (WHALE) wuggish (SPOON) feppish (CAR) talish (COMB)

2.1 Example Triad


2.2 Book Organization

Documents

Margaret Shavlik Thesis