37
1 Nature or Nurture? An Investigation of Demographic and Environmental Characteristics’ Impact on Infant Cognition Amanda Budow MMSS Senior Thesis Adviser Sue Hespos

Nature or Nurture? - Northwestern University

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Nature or Nurture? - Northwestern University

1

Nature or Nurture?

An Investigation of Demographic and Environmental Characteristics’ Impact on Infant Cognition Amanda Budow MMSS Senior Thesis Adviser Sue Hespos

Page 2: Nature or Nurture? - Northwestern University

2

Acknowledgements

There are many people without whom this paper would not have been

possible. First and foremost a special thanks to Professor Sue Hespos, without

whose research, support, and sense of humor this thesis would have been an

utter disaster. Secondly, to the research staff of the Northwestern Infant

Cognition lab for putting up with my endless research, analysis, edits, and

complaining. And lastly, a huge thank-you to my friends and family for all their

unwavering support.

Page 3: Nature or Nurture? - Northwestern University

3

Abstract

The body of research surrounding infant cognition has provided valuable

insight into the capabilities, limitations, and normal trajectory of early cognitive

capabilities. However, little investigation has been done to examine the

relationship between cognitive development and other innate characteristics.

The aim of the present analysis is to examine the relationship between early

cognitive development and environmental, socioeconomic, and demographic

characteristics. By means of regression analysis, the relationship between infant

cognition of substances and various characteristics was modeled. Results

showed that not only did no characteristic emerge as predictive of normal

cognitive capabilities, but also a composite of environmental and demographic

variables were not significantly related to cognition. Rather, random assignment

and habituation attributable to experiment design accounted for over fifty percent

of the variance in cognition.

Page 4: Nature or Nurture? - Northwestern University

4

Introduction

Over the past three decades, the body of research surrounding infant

cognition has grown tremendously. We have a better understanding not only of

infants' cognitive capabilities but also their limitations. Despite their inability to

communicate using language, infants can retain and interpret information from

the first days of life. Recent research has provided valuable insight into infant

cognition. For example, infants understand early principles of how objects

behave and interact (Spelke 1994), have a foundation laid for language

acquisition (Ferry, Hespos, Waxman 2010), can interpret spatial relationships

(Baillargeon 1998; Hespos & Piccin, 2009; Hespos & Spelke 2004), and

differentiate between solids and liquids (Hespos, Ferry & Rips, 2009). From

these studies and more we can now make scientifically backed claims regarding

infant knowledge.

But, like any other academic field, as the body of research grows new

questions arise. In this case, a field that was once based on assumptions of

naiveté is now filled with evidence of sophisticated concepts and a predictable

developmental trajectory. Like any other scientific claim, these assertions are

based on overarching patterns and trends. With such a large body of research in

this field, there is now an opportunity to look at infant cognition on the individual

level. We know that not every infant follows the exact same developmental

trajectory, but are there any underlying patterns in these individual differences?

More specifically do any explicit characteristics, whether demographic or

environmental, systematically affect cognitive capabilities? This question, which

Page 5: Nature or Nurture? - Northwestern University

5

alludes back to the age-old debate of nature versus nurture, has yet to be

explored.

The present analysis will investigate existing, significant research for

trends in cognition. The aim is to use regression analysis to perform a

preliminary search for any individual characteristics predictive of cognitive

capabilities.

Literature Review

The field of infant cognition is relatively new. Jean Piaget laid the

groundwork with his constructivist approach to infant cognition (1952; 1954).

However, the real hero to modern infant cognition research was Robert Fantz

with his introduction of visual preference techniques for studying infants (1958).

This method, which is described in detail later on, allowed for the discovery of

novelty preferences in infant looking patterns (Fantz, 1964). This paired with

Sokolov’s discovery of human’s orienting reflex, linking attention-related

behaviors to quality of memory and learning (Columbo & Mitchell, 2009). From

there, the research took leaps and bounds; making scientifically backed claims of

infants’ ability to perceive and remember objects and events. Today, the field

focuses on infants’ ability to develop and use internal mental representations.

Without going into too much historical detail, it suffices to say that we now have a

much better grasp on the human developmental trajectory from an early age.

With this maturing body of research come unavoidable implications. One

such consequence is widespread debate about the origins of knowledge. The

Page 6: Nature or Nurture? - Northwestern University

6

nature versus nurture, or nativism versus empiricism argument is central in most

cognitive research, and infant cognition is no exception. At the center of this

particular debate is the concept of object representation and whether infant

knowledge is innate, purely experience based, or somewhere in between.

Elizabeth Spelke leans toward the nature side of the argument, claiming that

perception according Gestalt’s theories of solidity and continuity is innate

(Spelke, 1994). Solidity principle states that one object cannot pass through

another, while continuity states that objects do not blink in and out of existence

but are continuous through time and space (Werthheimer, 1938). Thus, infants

reach predictably for moving objects and act surprised (look comparatively

longer) when an object violates one of these core principles.

The empiricist argument claims that after preliminary encounters with a

new object or category an infant forms a primitive “all or nothing” concept defined

by a rule. Then, with time and experience, he or she elaborates on this concept

by identifying increasingly refined variables that result in increasingly accurate

interpretations of how objects behave and interact (Hespos & Baillergeon, 2001).

Ultimately, this bottom-up processing mechanism refers back to a Piagetian

constructivist view. While the studies referred to above go beyond the cognitive

exploration done by Piaget, they nonetheless allude to knowledge constructed

through experience. This kind of bottom-up processing is also traditionally more

accommodating of change over time (Cohen, 2008).

At this point, it is a consensus among prominent researchers in the field

that infants can indeed categorize and form expectations of objects they

Page 7: Nature or Nurture? - Northwestern University

7

encounter. However, researchers are still investigating the origins of this

knowledge, and how it develops and adapts. This research brings up another

related inquiry: are all individuals on a more or less common trajectory when it

comes to these cognitive abilities? Returning back to the question of nature

versus nurture; do any particular innate characteristics systematically affect

infants’ ability to categorize objects and events at a common age? This

particular angle of the nature-nurture debate has yet to be pursued.

It has been established by prior research that young children and toddlers’

cognition skills are affected by innate characteristics such as race and gender, as

well as environmental factors such as socioeconomic status. For example,

according to a longitudinal study preformed by the U.S. Department of Education,

children’s test scores upon entering kindergarten vary significantly by

socioeconomic status. Specifically, cognition scores of high SES children are

60% above those of the lowest SES group (Lee & Burkam, 2002). Additional

studies have documented significant effects of gender, race and ethnicity, as well

as premature births on cognition. In the realm of race research, Jensen found a

Black-White group cognitive difference among three year olds (Peoples et al.,

1995). This same meta-analysis pointed to race differences observable at birth,

through brain size, and asserted that Whites, on average, have a larger head

circumference than Blacks (Rushton, 1993). On a final note, a meta-analysis

found significant cognitive differences among five year-old children born pre-

term; the mean cognitive scores of preterm-born cases and term-born controls

were directly proportional to their birth weight (R2 = 0.51; P<.001) and

Page 8: Nature or Nurture? - Northwestern University

8

gestational age (R2 = 0.49; P<.001; Bhutta et al., 2002). Thus, we see

meaningful evidence suggesting that at some point in development, these

demographic and environmental characteristics play a role in cognition. The

question, then, is when do these attributes begin to make an impact?

This investigation also makes another significant contribution to the

literature surrounding the topic of cognition. A major complaint within the

psychology field is that the majority of experiments are performed on extremely

limited subsets of the population yet subsequently generalized to the population

in its entirety. Thus, research performed on a limited sample becomes a human

psychological universal, defined as “core mental attributes that are shared at

some conceptual level by all or nearly all non-brain damaged adult human beings

across cultures” (Norenzayan & Heine, 2005). Many legitimate questions have

been raised questioning the integrity of these broad universality claims. First,

there is the complaint that all psychology subjects are “WEIRD,” a term coined by

Henrich in a new paper that describes research participants as westernized,

educated, industrialized, rich and democratic (2009). On a similar note, Jeffery

Arnett found that 67% of American psychology research dealt exclusively with

undergraduate psychology students. Furthermore, broadening the research to

the international level the portion of research preformed exclusively on

undergraduates increases to 80% (Arnett 2008). While undergraduates do

present an easy access subject pool it can easily be argued that they are a non-

representative subset of the population. Arnett goes as far as claiming that

“psychological researchers in the United States restrict their focus to less than

Page 9: Nature or Nurture? - Northwestern University

9

5% of the world’s total population,” thereby neglecting the other 95% (Arnett,

2008). Until recently, this trend in research has been largely overlooked.

However, a threat to standard sampling practices of this proportion could

threaten most research in the field.

As it pertains to cognitive and developmental psychology, one particular

subset of a population may develop cognitive capabilities in one way but others

may not, especially when you look outside of American culture. However, if the

result derived from the subset of the population is immediately generalized we

find ourselves making crippling faulty assumptions. Put simply by Paul Rozin,

“psychology produces research findings that implicitly apply to the entire human

population, the entire species. Psychological studies, journals and text books in

the United States describe the nature of social, emotional, and cognitive

functioning with the assumption that the processes described apply to all human

beings” (Rozin, 2006). While the research findings are undoubtedly significant

within the sample population, there could still be a dangerous error in

generalization. Making assumptions of universality could skew our

understanding of human psychological development, both on a cognitive and

social level.

Unfortunately, this generalization risk is difficult to test for considering

most psychology research opts for the convenience sample of “WEIRD”

American undergraduates. In most cases, the nature of the existing research

makes benchmarking typical sample populations against other potential

subgroups close to impossible. Thus, most psychology research implicitly

Page 10: Nature or Nurture? - Northwestern University

10

assumes that its investigations are “de facto universals.” Ara Norenzayan and

Steven Heine argue that “the bedrock of the psychological database, consisting

of cumulating layers of findings from Western middle-class college educated

young adults and their young children, prevents us from testing this assumption”

(2005). However, one possible approach is explored in this analysis with the

focus on infants. Not only are infants a divergence from the typical

undergraduate sample, but also they are neither educated nor democratic, and

have no personal wealth, eliminating three of the five “weird” characteristics

outlined by Henrich. Furthermore, at such a young age the impact of living in a

western industrialized society is likely a less significant effect, weakening the

impact of the remaining two “WEIRD” characteristics.

On the other hand, even the current sample is far from perfect. Like most

other psychology research the sample is limited by convenience. Participating

infants were largely from the Chicago and Evanston areas, and while there was

en even split between males and females for the most part, the sample did skew

toward higher socioeconomic status (proxied by parental education levels) as

well as certain race and ethnicity classifications, namely Caucasians. Thankfully,

the regression paradigm for analysis largely corrects for this kind of skew by

looking at trends on an individual participant basis.

Page 11: Nature or Nurture? - Northwestern University

11

Data Description

Source:

The data used in this analysis comes from the research of Professor

Susan J. Hespos at Northwestern University. The data set consists of looking

time and supplemental descriptive data for 267 healthy, full-term infants living in

and around the Evanston area. Data was collected voluntarily from participants in

two ongoing studies over the course of the past four years and aggregated for

the purpose of this analysis. The list of infants was obtained from a commercial

mailing list. Potential participants were contacted by letter and participants only

took part with parental consent. All studies use a habituation-dishabituation

model to examine infant ability to form and distinguish between categories.

Participants’ parents provided all supplemental data, including race,

ethnicity, gender, birth weight, gestation length, parental education levels, and

infant’s age. All information, excluding gender and age, was stored anonymously

and not paired with looking time data until the time of this archival analysis. While

age, gender, gestation length and birth weight were always provided, parents

were not required to provide race and ethnicity information or parental education

levels.

Original Research:

When investigating questions of infant cognition all researchers face a

common obstacle: young infants lack the ability to communicate verbally. So to

explore questions regarding infant cognition researchers often use a habituation-

dishabituation method, and the studies used in this analysis are no exception.

Page 12: Nature or Nurture? - Northwestern University

12

Because infants have good control over eye movements even at an early age,

these studies use looking time as a metric for interest in an object or event. The

present study examines two habituation-dishabituation studies, each with

multiple experiments, all of which were aimed to investigate cognitive

development. Specifically, one study investigated infants’ ability to form

expectations based on behavior of solids and liquids, while the other investigated

infant ability to distinguish between quantities of a sand substance.

In a habituation-dishabituation study, the infant is familiarized or

habituated to one set of objects or events by repeated presentation. All trials

were infant-controlled; that is to say that the each stimulus event was repeated

continuously until the infant ended it by either looking away or looking for a

minute total. In these studies, a participant was considered habituated after

showing a 50% decline in looking time between the first and last three

consecutive trials. Once this habituation criterion was met the study moved on to

test trials which alternated between presentation of a novel and familiar event. In

both studies, infants sat in a parents lap facing a wooden display stage where an

experimenter presented the events. The stage was equipped with cameras so

that research assistants could use live coding to record when an infant was

looking attentively at the on-stage stimuli. It is assumed that if an infant

understands and internalizes the concept presented in habituation trials, they will

be less interested in the familiar event in test trials. Thus, if an infant looks

longer at the novel event in test trials it follows that they have understood the

concept of interest.

Page 13: Nature or Nurture? - Northwestern University

13

Study 1:

The first experiment (Hespos, Ferry & Rips, 2009) investigated whether or

not five-month-old infants expect solids and liquids to behave differently. The

focus on expectations was a key attribute, as the goal of the experiments was to

determine whether infants can use physical cues of a solid or liquid to predict

later behavior of the same substance. Specifically, the stimulus was a rotating

transparent glass filled with either a blue solid or liquid substance of the same

quantity. By rotating the glass, the experimenter showcased the property of

surface movement. In test trials, two stimuli were alternately presented for six

total trials. In each of these trials either a blue solid or liquid was poured back

and forth between two glasses over eight-second intervals until the infant looked

away and the trial ended. If the participant habituated to the liquid event, the

solid test trial was considered novel and the liquid familiar (and vice versa in the

case that habituation was to solid).

In each of the experiments, the majority of participants had significantly

lower looking times for the familiar stimuli in the test trials. For example, in the

first experiment whose procedure is detailed above 26 of 32 total infants looked

longer at the novel event in test trials. Overall, average looking time was almost

ten seconds longer for the novel event regardless of which stimulus the

participant was habituated with. In test trials infants were surprised to see a solid

acting differently than the liquid event had they been habituated to the liquid

stimulus. Thus, it can be said that infants can distinguish between solids and

liquids.

Page 14: Nature or Nurture? - Northwestern University

14

The particular experiment described above tested infants ability to

generalize the property of surface movement demonstrated in the habituation

trials to test trial events. In the subsequent experiments in this study, various

other properties were tested among five month-old participants. Although the

habituation procedure remained the same the test trial displays demonstrated

other properties such as surface penetration and cohesion.

Study 2:

The second study's aim was to discover whether or not infants can

distinguish between quantities of sand. The participants ranged four different

age groups in order to track the developmental trajectory of the quantity

discrimination skill. The experiments used the same habituation-dishabituation

paradigm described above but this time the goal was to see if habituating

participants to a specific quantity of the substance would result in a novelty effect

in test trials when a new quantity was presented.

Infants were assigned to either the small or large quantity condition for

their habituation. In habituation trials, the infant watched as the experimenter

repeatedly poured blue sand from a cup onto a plate, and subsequently emptied

the plate off stage. The procedure was timed on a 15-second cycle and repeated

until the trial ended. Infants saw either the large or small quantity of sand (but

not both) in habituation trials. In test trials, the experimenter alternated between

the small and large pile stimulus for six total trials.

The participants belonged to one of four age groups: three, seven, ten,

and thirteen month-olds. Each was analyzed individually after data was collected

Page 15: Nature or Nurture? - Northwestern University

15

yielding significant yet varying results in each age group. Small to large quantity

ratios varied between a 1:4 ratio (tested only among 3,7, and 10 month olds) and

a 2:4 ratio (tested only among 7,10, 13 month olds). In the first experiment,

which tested differentiation between 1:4 quantity ratio, all age groups (3, 7, and

10 month olds) showed significant novelty preferences in test trials. An ANOVA

analysis across all three age groups produces a significant main effect, (F(1,76)

= 22.04, η2 = .23) (Hespos et a. 2010). In the subsequent experiment, the 2:4

quantity ratio was tested among seven, ten, and thirteen month-olds. The results

here were complicated by gender. While females showed a significant novelty

effect as expected, boys across all age groups showed a significant familiarity

effect. Nonetheless, both genders showed significant differential looking

between the two stimuli and thus results can still be reported successfully. Due

to the alternate trend among males, the present data set used reverse coding for

males in the second experiment. Thus, for effected male participants, a

familiarity effect signals comprehension of the concept and a novelty effect

signals defection.

Aggregation:

To form the data set used in the present analysis, results from both

studies were combined and organized by participant. All published results were

analyzed and reported prior to the beginning of this analysis. Original analysis

included mean looking times for novel and familiar test stimuli and overall

ANOVA. In both studies participants showed a significant increase in looking time

between the familiar and novel test trials, qualifying the study for inclusion in this

Page 16: Nature or Nurture? - Northwestern University

16

archival analysis. For the purposes of this analysis, subject information was

compiled in a customized spreadsheet which included both raw data for looking

times and supplemental descriptive data including age, gender, parental

education levels, race, ethnicity, birth weight and gestation length.

The total sample includes 267 infants, 152 male and 115 female. In the

water study, there were 107 total infants, 63 male and 44 female, with an

average of 29.5 participants in each of the four experiments. Of these

participants, more than 80 exhibited a significant bias in test trial looking time

toward the novel stimulus. The intended age for study participants was five

months; all participants were between 4.5 and 5.5 months old. In the sand study,

there were 160 total participants, 89 male and 71 female. The sand studies

tested infants of varying ages; specifically, all participants were three, seven, ten,

or thirteen months old.

Overall, the most common parental education level was “college or higher”

for both mothers and fathers. Earlier studies were skewed toward more highly

educated parents and white, non-Hispanic participants. The average participant

birth weight was 7.67 and gestation length was 39.5 weeks. To participate,

infants had to be considered fully mature, meaning they had to have a birth

weight above 5 lbs 8 oz and a gestation length longer than 37 weeks.

Variable Specification

Average difference in looking time = P3: This variable is a key metric in all of

the studies used in this data set. As explained previously, each of the

Page 17: Nature or Nurture? - Northwestern University

17

experiments consisted of a habituation period (6 - 9 trials) and a six trial test

period. In the test portion of the experiment the trials can be viewed in pairs,

alternating from novel to familiar or vice versa depending on the participant’s

random assignment. For each pair, the difference in looking time between novel

and familiar is recorded and averaged with the previous pair’s difference. At the

end of the test trials, three pairs’ differences will have been recorded and

averaged and this number is recorded as P3.

This variable ultimately describes whether the infant has successfully

internalized and distinguished the stimuli by appropriately interpreting the cues

presented in the habituation trials. If a participant’s P3 > 0, their average looking

time at the novel stimulus exceeded that of the familiar. It follows that in this

case, the infant categorized the cues from habituation, and used them to

distinguish between the two stimuli in test trials. Thus, this variable serves as the

dependent variable for the two OLS regressions in the primary analysis.

Dummy variable P: This variable is a simplified form of the P3 variable

presented above. Rather than use a continuous variable that can range from -60

to 60, P is a binary simplification of the P3 variable. Essentially, if the infant

looks longer on average at the novel stimuli their P would be 1, while those

participants who failed to look longer at the novel stimuli would have a 0

recorded. Specifically:

P = {0,1}. P = 1 if P3 > 0 and P = 0 if P3 <(=) 0.

Looking_Time_C1: The dependent variable from the original research. Total

test trial looking time at a specified stimulus measured in seconds, in this case

Page 18: Nature or Nurture? - Northwestern University

18

condition one (C1). For example, in the water study the liquid condition was

assigned C1, thus the Looking_Time_C1 variable would be total looking time at

the water stimulus in test trials. If the participant was also habituated to liquid,

we expect their Looking_Time_C1 to be relatively lower. On the other hand, if

the participant was habituated to solid we expect looking time to be longer.

Age: specifies each participants exact age, in months. It is calculated by the

following formula:

Age = # of whole months of age + # of additional days/ 30

Gender: records whether the participant is male or female. In the case of the

second Sand study, results were reverse coded by gender. Specifically:

Gender = {0,1}, where 0 = female, 1 = male

Parents’ education level: classifies the highest attained level of education for

each parent, organized by degree. Specifically:

Mom_EDU = {1,2,3,4} and Dad_EDU = {1,2,3,4}, where 1 = some

high school, 2 = high school diploma, 3 = some college, 4 = college

or higher

Race: In this analysis, race is organized into four categories specified by the U.S.

government. In this case, a fifth category for “mixed” because some candidates

were biracial. Participants were not required to provide this information, and

those who opted out were excluded from the analysis.

Race = {0,1,2,3,4), where 1 = asian, 2= black, 3 = pacific islander,

4= white, 5 = mixed

Ethnicity: defined as whether or not the participant was Hispanic, thus:

Page 19: Nature or Nurture? - Northwestern University

19

Ethnicity = {0,1}, where 0= Non-Hispanic, 1 = Hispanic

Birth weight: Infant’s weight on his or her birth date, as reported by the

consenting parent. Specifically calculated by the following formula:

Birthweight = Lbs + oz/16

Gestation_Length: number of weeks the mother carried the infant before

delivery, as reported by the consenting parent.

Study: specifies which study the participant took part in. Specifically:

Study ={1,0}, where 1 = Sand, 0 = Water

Total: represents the participant’s total looking time across both habituation and

test trials, measured in seconds. Specifically:

Total = Looking_Time_T1 + Looking_Time_T2 + Looking_Time_T3

+ Looking_Time_T4 + Looking_Time_T5 + Looking_Time_T6

Or, equivalently:

Total = Looking_Time_NOVEL + Looking_Time_Familiar

This variable was included as a proxy for attention span. Because it includes

total looking from both the novel and familiar test trials it is not expected to be an

indicator of whether the infant looked preferentially. However, due to the nature

of the P3, P, and Looking_Time_C1 variables the total looking time is certainly

related and exclusion of it would bias the overall analytical model.

Hab_C1: Binary variable representing the participant’s randomly assigned

habituation condition. Specifically:

Hab_C1 = {1,0}, where for SAND 1= large quantity, 0 = small

quantity and for WATER 1 = liquid and 0= solid

Page 20: Nature or Nurture? - Northwestern University

20

The Models

This analysis consisted of a primary and a complementary analysis, each

consisting of three separate regressions. The goal was to investigate any

systematic patterns in cognition attributable to demographic or environmental

characteristics. Thus, the primary analysis investigated the effect of these

characteristics (listed in variable form in the previous section) on the degree to

which the participant showed a novelty preference, if at all. As stated previously,

a novelty preference demonstrated by P3 > 0 implies that the participant was

able to internalize and distinguish the differences between stimuli. Each of the

three regressions examined the relationship between preferential looking time

and innate characteristics, first by looking at each respective study individually

with an ordinary least squares (OLS) regression and subsequently combining the

two to use a probit regression. Thus, the three models were as follows:

Water/ Sand:

P3 = ß0 + ß1Age + ß2 Gender + ß3 Mom_Edu + ß4 Dad_Edu + ß5 Ethnicity + ß6

Race + ß7 Birthweight + ß8 Gestation + ß9 Total +

All (probit model):

Pr(P = 1 | Age, Gender, Mom_Edu, Dad_Edu, Ethnicity, Race, Birthweight,

Gestation, Total, Study) = (X`ß) +

Where X represents the vector of innate characteristics listed above.

Page 21: Nature or Nurture? - Northwestern University

21

A key feature of the above regression models is that the study design is

only accounted for in the error term. In other words, the only specified variables

are innate demographic or environmental attributes. Thus, not only will each

variable’s resulting coefficient and t-value be valuable, but also the overall R2

metric will represent the portion of the variance in P3 accounted for by this

composite of descriptive variables. Thus looking time will be examined on a

basis irrelevant to condition, the model accounts for which condition they were in

independently of the analysis.

The creation of binary variable P makes it possible to use a probit

regression to analyze all of data as an aggregate set. While analysis of each

study individually is undoubtedly important, it is also advantageous to analyze its

aggregate for two reasons. Firstly, the combination makes for a larger data set

and thus more powerful results, secondly if any trend stands out within one study

it would be more credibly if it could be replicated in a different study of similar

design. If it could not be replicated, it would raise new questions of why cognitive

patterns appear in some studies but not others. Furthermore, the probit

regression uses a probability paradigm, exploring outside of a linear trend for a

relationship between looking time and innate characteristics. Thus, in the case of

this regression, any significant coefficients would account for a significant

increase in the probability that the infant looked longer at the novel stimulus. To

use the probit model it was necessary to impose the assumption of a normal

distribution onto the data set. While this may be a strong assumption it is

appropriate due to clustering around the mean for most of the descriptive data.

Page 22: Nature or Nurture? - Northwestern University

22

The complementary analysis, on the other hand, not only included the

descriptive variables from the first three regressions but also accounted for the

original study paradigm. Specifically, it included the independent and dependent

variable from the original experiment, namely habituation condition on the right

side of the equation and looking time specific to one condition on the left. Thus,

the regression measures the effect of a participant’s randomly assigned

condition, along with additional supplemental characteristics, on preferential

looking time. For example, in the Water studies, the dependent variable would

be total looking time at the liquid stimulus in test trials. In this same example the

right side of the equation would include all the environmental and demographic

variables as well as Hab_C1, coding whether or not the participant was

habituated to liquid as well. The liquid stimulus in the water study and the large

quantity stimulus in the sand study were randomly assigned as C1, while solid

and the small quantity were assigned C2. Using the same variables, an OLS

regression was run on each of the studies individually as well as on an

aggregation of the two. Each regression fit the following OLS model:

Looking_Time_C1 = ß0 + ß1Age + ß2 Gender + ß3 Mom_Edu + ß4 Dad_Edu + ß5

Ethnicity + ß6 Race + ß7 Birthweight + ß8 Gestation + ß9 Total + ß10 Hab_C1 +

Results

The primary analysis revealed that no single demographic or

environmental characteristic consistently predicted the variation in the average

Page 23: Nature or Nurture? - Northwestern University

23

difference in looking time between novel and familiar stimuli. Additionally, all

estimated coefficients were low in magnitude (less than two in absolute value), a

looking time differential considered inconsequential. Furthermore, all the

supplemental variables of interest (namely gender, race, ethnicity, and parental

education levels) in each of the three regressions accounted for no more than

five percent of the total variation in looking time.

In the Sand and Water OLS regressions using P3 as the dependent

variable only a small portion of the variation could be attributed to the variables of

interest, R2 = .046 F(9, 109) = .58, p > .005 and R2 = .038 F(9, 100) = .45, p >

.005 respectively. Thus, at most, only 4.6% of the variation in looking time could

be attributed to variables of interest, and only inconsistently. Furthermore, as

demonstrated in Table 1, not only did no single coefficient have a high magnitude

coefficient (most were fractions), but also none were statistically significant with p

< .05. Even by relaxing the threshold for statistical significance as high as p < .1

no variable proved significant in either case. Similarly, the aggregate analysis

using the probit model had an even lower pseudo R2, allowing the model to

account for only 2.19% of the overall variation. Table 2 specifies the exact

outcomes of the aggregated probit regression. Just as in the previous OLS

regressions, no one variable had significantly high magnitude or approached the

p = .05 threshold.

However, in all three regressions some variables did stand out as they

approached significance. Study, for example, had the lowest p-value in the

aggregate regression with ß = -.34, z(269) = -1.92, p =.05. Although it was not a

Page 24: Nature or Nurture? - Northwestern University

24

high impact attribute, study design does predict variations in looking time more

reliably than other variables. Additionally, in both study-specific regressions the

birthweight variable had the highest t-statistic in terms of absolute value, and

thus the lowest p-value. Although the value was not high enough to prove that

the slope of the estimated regression differs significantly from zero, it does stand

out. Furthermore, neither coefficient is large enough to be considered a major

effect, ß = 1.37, t(119) = 1.14, p = .25 and ß = -1.72, t(110) = -1.49, p = .14 for

Sand and Water respectively. A two second change in average looking

differential would likely not shift overall trends.

In the complementary analysis, on the other hand, the R2 values

increased drastically. For instance in the OLS regression performed on the

Water study the R2 value improved by almost 60% and proved statistically

significant; specifically, R2 = .632 F (10, 107) = .16.48, p < .001. The original

independent variable Hab_C1, which in this case referred to whether or not the

participant was habituated to liquid, was high in magnitude and significant, ß =

34.95, t(107) = 5.42, p < .001. Unlike in the primary analyses where variable

manipulations were estimated to cause no more than a five second change in

looking time, the habituation condition accounts for more than a thirty-second

difference. This big of a change would likely switch overall looking preference

from one test stimulus to the other, a claim supported by the study’s original

ANOVA analysis. Furthermore, none of the demographic or environmental

variables of interest showed a dramatic increase in coefficient magnitude or

statistical significance. In fact, as can be seen in Table 3, none of the other

Page 25: Nature or Nurture? - Northwestern University

25

variables except for Total (ß = .30, t(107) = 11.17, p < .001) approached

statistical significance. Considering that total looking time overall is logically

correlated to looking time within a particular condition, it is not surprising that

Total gained statistical significance. However, it is also intuitive that the

magnitude of its coefficient be relatively small, as it should not predict preferential

looking time, only looking time in general.

In the Sand complementary analysis similar results held, with R2 = .590

F(10, 148) = 21.27, p < .001. Thus, adding the Hab_C1 variable caused a

corresponding 55% increase in accuracy of the model. Specifically, a change in

the participant’s habituation condition from large to small (Hab_C1 (ß = 7.38,

t(159) = 1.85, p < .05) was estimated to result in an almost eight second looking

time differential. Although this coefficient was smaller in magnitude than the

aggregate and Water results, it was nonetheless the largest within the

complementary Sand analysis. On a different note, in the complementary Sand

analysis both birthweight and gestation approached statistical significance (p =

.037 and p = .041 respectively).

In the aggregated complementary analysis the OLS regression showed an

even stronger result, both the R2 and F value increased significantly, with R2 =

.616 F(11,254) = 36.97, p < .001. Furthermore, as demonstrated in Chart 1, the

R2 increase marked a 60% increase from the pseudo R2 of the aggregate probit

analysis. Similarly, the Hab_C1 variable was of large magnitude and significant,

ß = 18.88, t(254) = 5.27, p < .001. This result is consistent with the respective

ANOVA’s published with the original research and signifies the reliance of

Page 26: Nature or Nurture? - Northwestern University

26

looking time on habituation condition. In addition, both the Study and Total

variables gained statistical significance in this regression, with ß = -21.33, t(254)

= -55, p < .001 and ß = .26, t(254) = 16.52, p < .001 respectively. The Total

result fits with the finding in the primary analysis and can be accounted for by the

same logic. The Study coefficient however follows directly from the data.

Overall, infants looked longer in the Water study relative to the Sand study,

across all conditions and phases (habituation and test) of the experiment. Thus,

considering the Study value decreases to the binary zero option in the case that

the infant participated in the Water study, it is not surprising that it negatively

correlated with a metric of looking time. Nonetheless, this means that which

study a participant took part in does account for some of the variance in overall

conditional looking time.

While none of the variables of interest gained statistical significance in the

aggregate OLS regression, many of them changed marginally in magnitude (see

Table 4). For instance, in addition to the previously discussed Hab_C1, Study,

and Total variables, both Gender and Ethnicity had coefficients with magnitudes

greater than one. Specifically for Gender ß = 3.0, t(266) = .82, p = .41 and

Ethnicity showed ß = 5.96, t(266) = .1.07, p = .29 in the aggregate OLS

regression. While a three to five second increase in looking time may not be

consequential in the grander scheme of things it is nonetheless interesting to

note. It presents the possibility that subjects of different gender or ethnicity may

systematically look longer at certain stimuli. In the complementary analysis

coefficients also show greater variation between studies. For example, the

Page 27: Nature or Nurture? - Northwestern University

27

Gender variables changes from a magnitude of 5.53 in the Sand study to a

negative coefficient equal to -1.88 for the Water data. Similarly, the coefficient on

Ethnicity also varied from 4.36 to 8.7 seconds. While this, in combination with

the significance of the Study variable, suggests that different characteristics

affect trends in looking time differently in each study, the variables lack statistical

significance.

Implications

The implications of the preceding analysis and results are three-fold.

First, the above analysis asked a novel question within the field of developmental

psychology, specifically infant cognition. The fact that none of the later influential

demographic and environmental characteristics examined showed any significant

effect on differences in infant looking time is valuable in itself. It implies that at

least within the category of substance knowledge, no intrinsic bias consistently

alters the typical cognitive development trajectory of infants. While there may be

other cognitive skills that develop differently in individuals of varying race or

different gender, the skills approximated here were unaffected. Ultimately, the

drastic change in R2 values paired with the repeated significance of the Study

variable implies that experiment design and deliberate manipulation of the

dependent variable (Hab_C1 in this case) account for the majority of the variation

in looking time.

Secondly, to reference the title of this paper, the results approach the

question of nature versus nurture, or nativism versus empiricism. While we

Page 28: Nature or Nurture? - Northwestern University

28

cannot address the bigger question of where infant’s knowledge originates, we

can speak to how it varies. Overall, the results point toward the nurture side of

the debate. As documented in the literature review, many studies have found

significant effects of demographic and fixed characteristics such as

socioeconomic status, gender and race on cognition at a later age. Although the

cognitive skills tested in other studies are not a perfect parallel to the substance

studies investigated in the present analysis, they nonetheless point to an active

role for demographic characteristics later on in development. The fact that any

indication of these effects is absent from this particular study suggests that

environmental factors, or the “nurture” part of the equation, may play a larger role

in shaping differences in cognition.

This assertion requires a logical jump, assuming that the cause for later

cognitive differences are more likely a result of context, following from the

present result that they have no impact early on when “nature” characteristics are

already in place. Another possibility is that innate characteristics begin to play a

part in cognition later on. In this case while the above result would still be

significant, it would not apply to the nature versus nurture debate. While this is

only one small piece of a larger developmental puzzle it nonetheless makes a

significant contribution. Even within a cognitive category, the claim that neither

gender, nor socioeconomic status or race consistently effects cognition is a

strong one. It implies that variation in cognition of substances should more likely

be attributed to individual differences such as temperament and exposure.

Page 29: Nature or Nurture? - Northwestern University

29

Thirdly, these results strengthen the argument that psychological

researchers should hesitate before generalizing results to entire populations.

As stated in the literature review, many published psychology studies imply

generalization to larger populations than the one sampled. Broad universalities

often lead to error in understanding of human nature, both cognitively and

emotionally. For instance, were someone to generalize the above result to the

entire population we would assume that all cognitive skill sets (beyond just

substances) are categorically unaffected by factors such as race, gender and

socioeconomic status. However, this is certainly not the case. Not only is there

well-documented evidence that gender and race do play a role in cognition later

on, but also in one of the specific studies used here gender played a significant

role (as outlined in discussion of Study 2). Ultimately, by stepping outside of the

usual “WEIRD” sample common to most psychology studies we found a new and

interesting result. This suggests that research results derived from atypical

samples are valuable not only for the results themselves, but as a comparison to

similar research done on other samples. On the same note, research done on

undergraduates or other “WEIRD” samples should be wary of universality and

furthermore should be sure to elaborate on sample limitations in their discussion

of results.

Limitations of the Analysis

Like other studies, the scope of this analysis was somewhat limited by the

constraints of the sample. Ideally, this kind of analysis would be performed on a

Page 30: Nature or Nurture? - Northwestern University

30

perfectly diverse sample, with an equal representation of each race, ethnicity,

and parental education level. Unfortunately, a data set of that nature is nearly

impossible to find, due to location and convenience. The accessible subject pool

for psychology research includes mostly highly educated, white participants and

the field of infant cognition is no exception. Although any significant change in the

results would be surprising, a more diverse sample would be preferable and

results should only be generalized with caution.

Given additional resources, a study of this nature would ideally explore the

longitudinal trends in the relationship between demographic and environmental

factors and cognition. As stated previously, other research has found a

significant relationship between cognition and factors such as race,

socioeconomic status and gender yet the timeline of these effects is unknown.

By tracking individuals from a socioeconomic and racially diverse sample in

various cognitive tasks over time one could gather more concrete evidence for

when these innate factors do play a role. As the field of infant cognition expands

a meta-analysis of these kinds of studies may become possible.

Conclusion:

This analysis is an especially unique one because a lack of significant

results is equally, if not more fascinating as the discovery of a significant

correlate. Thus, the most striking result from this analysis is that no

characteristic variable has stood out as a significant predictor of infant ability to

distinguish between stimuli. Thus, we can indeed claim that at least in the realm

Page 31: Nature or Nurture? - Northwestern University

31

of this research, innate characteristics are not causing any significant bias in

cognition. Because later research has come to the opposite conclusion this

suggests that at some point after one year innate characteristics begin to play a

more profound role in the cognitive developmental tract. However, for the

purposes of this analysis, no attribute proved significant and overall variation in

cognition was largely attributable to study design.

Page 32: Nature or Nurture? - Northwestern University

32

Works Cited:

Arnett, Jeffery. (2008) The Neglected 95%: why American psychology needs to

become less American . American Psychologist, 63(7), 602-614.

Baillargeon, R. (2004). Infants' Physical World. American Psychology Society,

13(3), 89-94.

Bhutta, A.T. et al (2002). Cognitive and behavioral outcomes of school-aged

children who were born preterm. JAMA,288, 728-737.

Cohen, L. (2008). The Evolution of Infant Cognition: A Personal Account.

International Conference on Infant Studies.

Columbo, J. & Mitchell, W (2009). Infant Visual Habituation. Neurobiology of

Learning and Memory, 92 (2), 225-234.

Ferry, Hespos, Waxman (2010). Categorization in 3- and 4-month old infants; an

advantage in words over tones. Child Development, 81(2), 472-479.

Henrich, J. (2009). The weirdest people in the world. Behavioral and Brain

Sciences, in press.

Hespos, S. J. & Baillargeon, R. (2001). Reasoning about containment events in

very young infants. Cognition, 78, 207 – 245.

Hespos, S. J. & Baillargeon, R. (2006). Decalage in infants' reasoning about

occlusion and containment events: Converging evidence from action

tasks. Cognition, 99,B31-B41.

Hespos, S. J., Dora, B., Rips, L. J., & Christie, S. (2008). Comparing across two

piles: Further evidence of infants' quantity discriminations for substances.

Page 33: Nature or Nurture? - Northwestern University

33

Hespos, S.J., Ferry, A., Rips, L. (2009). Five-month-old infants have different

expectations for solids and substances. Psychological Science, 20(5), 603

– 611.

Hespos, S. J. & Spelke, E. (2004). Conceptual Precursors to Language. Nature ,

430, 453-456.

Hespos, S.J. & Piccin, T. (2009). To generalize or not to generalize: spatial

categories are influenced by physical attributes and language.

Developmental Science, 11(6), 951-958.

Hyde, J. S. (1981). How large are cognitive gender differences? A meta-analysis

American Psychologist, 36(8), 892-901.

Lee, V. E. & Burkam, D.T. (2002). Inequality at the starting gate. Economic Policy

Institute, Sept 2002.

Norenzayan, A & Heine, S.J.(2005). Psychological Universals- What are they

and how can we know? Psychological Bulletin, 131(5), 763–784.

Piaget, J. (1952) The origins of intelligence in children. New York, NY: W W

Norton & Co. 419 pp.

Rushton, P.J. & Jensen, A.R. (2005). Thirty years of research on race

differences in cognitive ability. Psychology, Public Policy, and Law, 11(2),

235-294.

Spelke, E. (1995). Intial Knowledge: Six Suggestions. Cognition on Cognition

(433-445).

Spelke,E. (1988). Nativism, Empiricism, and the Origins of Knowledge. Infant

Behavior and Development, 21(2). 17- 31.

Page 34: Nature or Nurture? - Northwestern University

34

Van Hulle, C.A. & Lemery-Chalfant, K. (2007). Genetic and environmental

influences on socio-emotional behavior in toddlers. Journal of Child

Psychology and Psychiatry 48(10).

Wilcox, T. & Baillargeon, R. (1998). Object individuation in infancy: the use of

featural information in reasoning about occlusion events. Cognitive

Psychology, 37, 97-155.

Page 35: Nature or Nurture? - Northwestern University

35

Appendix:

Chart 1:

Table 1: OLS Primary Analysis

Water Sand

Variable Coefficient T-Value Coefficient2 T-Value2

Age 1.84 0.74 -0.02 -0.07

Gender 1.13 0.52 -0.12 -0.06

Mom_Edu -0.31 -0.12 -2.16 -1.18

Dad_edu 0.12 0.04 0.43 0.26

Ethnicity -1.80 -0.45 2.40 0.74

Race -0.22 -0.16 0.03 0.02

Birthweight -1.72 -1.49 1.37 1.14

Gestation 0.17 0.16 -0.61 -0.69

Total 0.01 0.67 0.01 1.28

2% 4% 5%

62% 63%59%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

All Water Study Sand Study

Primary R2

Complementary R2

Page 36: Nature or Nurture? - Northwestern University

36

Table 2: Primary Probit Analysis

Variable Coefficient Z-Stat P>|z|

Study -0.34 -1.92 0.06

Age 0.03 0.91 0.36

Gender 0.14 0.86 0.39

Mom_Edu -0.08 -0.51 0.61

Dad_Edu 0.04 0.24 0.81

Ethnicity -0.14 -0.54 0.59

Race 0.10 1.07 0.29

Birthweight 0.00 0.03 0.03

Gestation 0.05 0.63 0.63

Total 0.00 0.01 0.92

Constant -1.60 -0.60 0.55

*z > 1.96 implies statistical significance, significant variables are bolded

Table 3: Complementary OLS Analysis

Water Sand

Variable Coefficient t-value Coefficient t-value

Hab_C1 34.95 5.42 7.38 1.85

Age -1.78 -0.23 -0.27 -0.44

Gender -1.88 -0.29 5.53 1.38

Mom_Edu 13.71 1.74 -4.39 -1.22

Dad_Edu -9.34 -1.14 -1.56 -0.46

Ethnicity 8.7 0.73 4.36 0.77

Race -4.35 -1.07 0.61 0.26

Birthweight -5.076 -1.48 4.95 2.1

Gestation 2.602 0.81 -3.57 -2.07

Total 0.2958 11.17 0.24 13.27

* t-value > 1.96 implies statistical significance, significant variables are bolded

Page 37: Nature or Nurture? - Northwestern University

37

Table 4: Complementary Aggregate Analysis

Variable Coefficient T-Value

Hab_C1 18.879 5.27

Study -21.33 -5.57

Age -0.12 -0.17

Gender 2.99 0.82

Mom_Edu 0.011 0

Dad_Edu -0.196 -0.06

Ethnicity 5.95 1.07

Race -1.52 -0.69

Birthweight 1.537 0.79

Gestation -1.59 -0.97

Total 0.26 16.52

* t-value > 1.96 implies statistical significance