Identifying Authors by Phonoprints in Their Characters' Names: An Exploratory Study

Abstract

If authors put words together in ways that can be recognized as wordprints (Hilton, 1990; Morton, 1979; Archer et al., 1997), do they put sounds together in identifiable ways when they invent names? Could they have unique sound prints (phonoprints) as well? This exploratory study compared phonemic patterns of fictional names in the poorly written Manuscript Story by Spalding and the extremely well-written Lord of the Rings and related works by J. R. R. Tolkein with names from an authentic public record, the nineteenth-century US Census. Phonotactic probabilities were determined using a calculator (Vitevitch and Luce, 2004) available on the Internet. When multivariate patterns of mean phonotactic probabilities at each ordinal phoneme position were considered, phonoprints emerged that merit further examination.

Full text

© American Name Society 2013 DOI 10.1179/0027773813Z.00000000048
names, Vol. 61 No. 2, June 2013, 101–21
Identifying Authors by Phonoprints
in Their Characters’ Names:
An Exploratory Study
Brad Wilcox, Bruce L. Brown, Wendy Baker Smemoe,
Sharon Black, and Justin Bray
Brigham Young University, USA
If authors put words together in ways that can be recognized as wordprints
(Hilton, 1990; Morton, 1979; Archer et al., 1997), do they put sounds
together in identifiable ways when they invent names? Could they have
unique sound prints (phonoprints) as well? This exploratory study compared
phonemic patterns of fictional names in the poorly written Manuscript Story
by Spalding and the extremely well-written Lord of the Rings and related
works by J. R. R. Tolkein with names from an authentic public record, the
nineteenth-century US Census. Phonotactic probabilities were determined
using a calculator (Vitevitch and Luce, 2004) available on the Internet. When
multivariate patterns of mean phonotactic probabilities at each ordinal
phoneme position were considered, phonoprints emerged that merit further
examination.
keywords author identification, phonoprints, fictional character names,
authentic names, nineteenth-century census, J. R. R. Tolkein, phonotactic
probabilities
An elementary school teacher noticed that children reading the Harry Potter fantasy
series were writing their own fantasies, with imaginative characters, fantastic settings,
and strange names. Some students shortened or expanded familiar names; others
compounded two known words to make a new name. Each child seemed to be
settling into his or her recognizable pattern.
Fiction authors are more sophisticated than children, but do they also have their
own individual patterns and preferences when they invent names? Do these patterns
differ from those of other authors or those found in authentic names?
Some linguists believe that each author creates an individual and identifiable word-
print by the way he or she puts words together (Hilton, 1990; Morton, 1979; Archer
et al., 1997). Although they are not as unique as fingerprints and are sometimes

102 BRAD WILCOX et al.
tentative and difficult to define (Croft, 1981), wordprints can be used to determine
probability of a writer’s identity and are used regularly in verifying document author-
ship (Baily, 1979; Holmes, 1994; Zheng et al., 2003; Iqbal et al., 2010). If authors put
words together in ways that can be recognized, might they also — consciously or
subconsciously — put sounds together in consistent individual patterns when they
invent words? Could each have his or her own sound print (phonoprint) as well?
Vitevitch et al. (1997) reported a significant correlation between what they called
phonotactic probability — the chance of certain sounds coming in certain positions
in English — and what they labeled neighborhood density — sequences of sounds
commonly found in close proximity. They found that consonant-vowel-consonant
content words comprised of common sounds tend to have many predictable lexical
neighbors. For example, some of the “neighbors” for the word cap would be cat, can,
cost, and tap. Another example which comes from historical Germanic and Latin
roots would be that /k/ is often followed by /w/ as in queen or quick. When authors
create fictional names, do they unwittingly create them from similar phonotactic
neighborhoods? Do they group sounds in typical ways or pair non-typical sounds
consistently? Regardless of varied backgrounds and skill levels, do authors of
fictional names use phonemic patterns that differ from those seen in nonfiction names?
This study was undertaken to explore such possibilities.
Purpose
Whether names are authentic, adapted, or created, processes of generating them have
been studied for years (Francis, 1981). However, few have examined the process
author by author. The purpose of this study was to analyze and compare the phone-
mic patterns of the male names found in a little-known manuscript by Solomon
Spalding (considered by many to be poor fiction), and The Lord of the Rings and
related works by J. R. R. Tolkein (considered by many to be excellent fiction) with
authentic male names from a public source: the US Census records of the nineteenth
century, the time at which Spalding produced his manuscript. Male single-word
names were used because of their prevalence in the selected texts. Recognizing
the broad diachronic comparisons between works attributed to the nineteenth and
twentieth centuries, we have considered this study as merely exploration to determine
if differences between authors and between fictional and authentic names merit
further investigation.
Phonotactic research
Typical methodologies for name study include structural analysis and contemporary
or historical comparison — allowing researchers to determine whether a word is part
of a specific language (Wu, 2010). Some researchers ask native speakers to confirm
whether a word “sounds” like the language or not (Young, 2004); others use a corpus
of a particular language for historical linguistic analysis (Downey et al., 2008).
None of these methods was completely adequate for studying the names in our
selected works. Traditional structural analysis of the names would require data not
currently available. Comparing the names to words in other languages would be

103
IDENTIFYING AUTHORS BY PHONOPRINTS IN THEIR CHARACTERS’ NAMES
tentative, since both novelists invented the languages from which character names
supposedly came; thus native speakers would not exist.
This study focused on phonotactic probability, previously defined as the general
frequency of occurrence of phonological segments and sequences of segments in a
given language (Jusczyk et al., 1994). For example, the vowel sounds found at the
beginning of the words eat and if are more common in English than the vowel sounds
found at the beginning of alms and oink. Similarly, consonant sounds such as those
found at the beginning of love, kiss, ton, and new are all much more common in
English than those found at the beginning of young and whip (Kessler and Treiman,
1997).
All sounds (phonemes) and pairs of sounds (biphonemes, labeled bifones by
Vitevitch and Luce [2004]) have varying levels of probability in English, depending
on their position in any given word or name regardless of origin. For example, Adam
and Solomon both have their origins in Hebrew, yet the average probability that in
English all the phonemes and bifones would come in the order they do is higher for
Solomon (Phonemes = .0716; bifones = .0081) than for Adam (Phonemes = .0231;
bifones = .0020). The probability that in English the sounds in the word Solomon
would come in the position they do is about 7%, and the probability of the pairs of
sounds found in Solomon coming in their ordinal positions is less that 1% on average.
In contrast, the sounds in Adam have a 2% probability of coming in the order
they do, with the pairs of sounds having less than two-tenths of 1% probability of
occurring. Phonotactic probability does not measure how common a word or name
is. There may be more men named Adam than Solomon. Phonotactic probability
deals only with prevalence of the sound sequence.
These determinations can be made using the probability calculator developed by
Vitevitch and Luce (2004), available on the Internet (<http://www.bncdnet.ku.edu/
cgi-bin/DEEC/post_ppc.vi>), which compares each word to the Brown Corpus,
a frequency database of standard American English created by Kucera and
Francis (1967). Each word can be entered phonemically or phonetically using the
computer-readable transcription method called Klattese developed by Dennis Klatt
(<http://129.237.66.-221/VLbrmic.pdf>). In this study, words were entered phonemi-
cally because little is known about the context-conditioned variations of many of
the names on a phonetic level. Klatt uses keys available on a keyboard to represent
unique sounds. For example, Edgar was entered as Edgx, Erchamion was rendered as
xCamian, and Borthand was borT@nd. The output of the calculator contains the
position-specific probability for each phoneme, the position-specific probability for
each bifone, and the sum of all phoneme and bifone probabilities.
Phonotactics, the body of rules that determine the constraints on the location of
sounds in the syllable structure of languages, has been used extensively to study
patterns in English names (Whissell, 2001; Shih, 2012) including the differences
between male and female names and ways phonological cues can predict gender
(Wright, 2012; Starr, 2012). Phonotactic probabilities have been used in relation to
language, memory, speech, and hearing. For example, researchers have examined the
prominence of certain beginning and ending sounds in words and syllables, exploring
how such word structures influence spoken-word recognition (Vitevitch, 2002;
Vitevitch and Luce, 1999). Others have looked at similarities in certain groups of

104 BRAD WILCOX et al.
words (Bailey and Hahn, 2001) relevant to language learning and development
(Storkel, 2001; 2003). Studies have examined speech errors and phonotactic con-
straints (Dell et al., 2000), as well as infants’ sensitivity to the sound patterns of native
language words (Jusczyk et al., 1993).
Only a few studies have examined invented words, including one demonstrating
that infants of similar ages could discriminate which non-words contained sounds
more common or less common in their native languages (Jusczyk et al., 1993).
Another explored the processing of non-words by deafened adults with cochlear
implants (Vitevitch et al., 2002). We know of no other studies of phonotactic
probababilities provided by the Phonotactic Calculator having been applied to
fictional or authentic name sources. Thus we are using an established method to
perform a new analysis.
Three name sources
The three names sources for this study were a manuscript by little-known nineteenth-
century author Solomon Spalding, The Lord of the Rings and related works by J. R.
R. Tolkein, and the US Census records of the nineteenth century. The Spalding
manuscript, dismissed by many as poor fiction, was included for its many author-
invented names that are not found or used elsewhere. Tolkein’s novels, on the other
hand, are generally recognized as examples of excellent fiction and were included for
their unusual names.
Names from the Spalding manuscript
Around 1800 Solomon Spalding (or Spaulding; 1761–1816) wrote a little-known fic-
tional story titled Manuscript Story (also referred to as Manuscript Lost or Manu-
script Found), a fictional account of a lost civilization of Native Americans who built
earth mounds in central and eastern United States. A group from Rome, traveling to
England around the time of Christ, were blown off course and landed on the North
American continent where they interacted with these native people.
Spalding’s fictional work includes unusual place names (e.g., Tolanga) and groups
of people (e.g., Sciotans and Kentucks). Common personal names like Tom are used,
but we identified 61 personal names not found elsewhere for use in this study (e.g.,
Lobaska, Bombal, Lamack, Helicon).
Pronunciation of names
Spalding did not provide a pronunciation guide for Manuscript Story, so two differ-
ent lists of pronunciations were used in this study: one based on general decoding
guidelines for English (Eldredge, 2005), a second based on decoding guidelines for
Latin, since the story’s travelers migrated from Rome.
An ANOVA analysis revealed that the English and Latin pronunciations differed
significantly in phoneme probability, the English pronunciations being more like stan-
dard American English in the Kucera and Francis corpus (1967), both for phonemes
(.289 compared to .220, F[1, 486] = 168.51, p < .0001, R2 = .258) and for bifones (.017
compared to .009, F[1, 364] = 134.79, p < .0001, R2 = .270). However, considering
the patterns of probabilities for each of the four ordinal positions of phonemes and

105
IDENTIFYING AUTHORS BY PHONOPRINTS IN THEIR CHARACTERS’ NAMES
bifones showed the two pronunciations virtually identical (see Figure 1); thus we
included only English pronunciations in the study.
Ordinal position comparisons are problematic across names of different lengths;
thus we used a telescoping design including four analyses: one for all names of four
or more phonemes (122 paired English and Latin names, analyzed for four ordinal
positions), one for all names of five or more phonemes (106 paired names, analyzed
for five ordinal positions), one for all names of six or more phonemes (72 pairs), and
one for all names of seven or more phonemes (22 pairs). Too few names had eight
phonemes to provide analysis. The results of the three latter analyses were consistent
with those from the analysis of all four-or-more-phoneme names.
Phonotactic descriptions
While all 61 personal names in the Spalding manuscript were included in the pronun-
ciation analysis, female names were not used in the remainder of this study; thus
we examined 55 male names by Spalding. The Spalding name containing the most
phonemes was Drafolick, with a total of eight. Three names containing only four
phonemes, Como, Droll, and Kato. Sambol and Baska, were the most like standard
English in the Brown Corpus, with average phoneme probabilities of .0712 and .0666.
Ulipoon was the least like English in the Brown Corpus, with an average phoneme
probability of .0330. Considering bifones, the name most like English in the corpus
was Hamul, with an average bifone probability of .0070. The name least like standard
English was Boakim, with an average bifone probability of .0010. Overall, the longer
the name, the less it was like standard American English.
Names from The Lord of the Rings
The Lord of the Rings trilogy, composed of The Fellowship of the Ring, The Two
Towers, and The Return of the King, was first published by J. R. R. Tolkien (1892–
1973) between 1954 and 1955. The Hobbit was published in 1937. After Tolkien’s
death a related collection of legends and tales about pre-middle earth was published
figure 1 A comparison of English pronunciations and Latin pronunciations of the Spalding
names in their phoneme probabilities at each of the first four ordinal positions.

106 BRAD WILCOX et al.
in 1977 called The Silmarillion. Some claim the entire story reflects Tolkien’s interest
in Germanic and Celtic mythology and folklore (Chance, 2004). These works intro-
duced civilizations and cultures with different languages and numerous unusual
names. For ease of reading in this study, all Tolkien’s works set in middle earth were
labeled with one title: The Lord of the Rings.
Tolkien, a linguist who became a professor of Anglo-Saxon in 1933, was well
versed in many languages, including medieval and modern Germanic and Celtic
tongues. With his knowledge and love of languages, he created several artificial
languages (based mostly on languages like Latin, Welsh, Finnish, and Old Norse);
vestiges of these invented languages appear in his fiction.
Character names in The Lord of the Rings were derived from several sources.
Tolkien explained that many of the names were from invented Elvish languages
(e.g., Legolas, Elrond, and Galadriel); others were translated from these languages
into English equivalents (e.g., Treebeard, Skinbark, Leaflock). Some were derived
from Old English to appear ancient (e.g., Theoden, Eomer, Erowyn). Some seem to
have been derived from the author’s childhood experiences, such as Sam’s surname
Gamgee, a word used for a cotton mill in the small town of Sarehole where Tolkien
grew up.
Pronunciations of names
Pronunciations for names in The Lord of the Rings were based on pronunciation
transcriptions from audio recordings of Tolkien reading his own books (<http://
inogolo.com/guides/lord+of+the+rings>). Pronunciations used in the film trilogy
directed by Peter Jackson were also consulted.
Phonotactic descriptions
This study used 197 male names from Tolkien that are not found in other sources,
selected from an Internet list (<http://lotr.wikia.com/-wiki/List_of_characters>) and
verified using the online Encyclopedia of Arda (<http://www.glyphweb.com/arda/>).
As few surnames or titles appear in the text, only first names were included. Names
given to two or more characters were used only once in the study.
The name with the most phonemes in The Lord the Rings was Celebrimbor, with
11. Oin had the least, with two phonemes. Among the names that were the most like
standard American English in the Brown Corpus were Sauron, Saruman, and Beren,
with average probabilities of .0792, .0723, and .0715 respectively. The least like Eng-
lish in the corpus was Azog, with an average phoneme probability of .0098. Focusing
on the bifones, the name that was most like standard English was Indis, followed
closely by Sauron and Saruman, with average bifone probabilities of .0126, .0111, and
.0101 respectively. The names least like English in the corpus were Oin, Azog, and
Olwe, with average bifone probabilities of .0001, .0002, and .0003. Names that were
longer were more like standard American English.
Nineteenth-century names
Throughout the 1800s both male and female given names were gathered and reporte d
on the US Census, but, to be consistent with the prevalence of male single-word
names in the books, only the 100 most common male first names on the census were

107
IDENTIFYING AUTHORS BY PHONOPRINTS IN THEIR CHARACTERS’ NAMES
used (<http://users.erols.com/dgalbi/names/us200.htm>; see also Erickson et al.,
2008). Their origins were identified using the Dictionary of First Names (Hanks and
Hodges, 1995). Although English is a Germanic language and name similarities and
overlaps occur, the origins provided in the dictionary were used with no alterations.
Origins of names
In the United States in the nineteenth century, the 10 most popular male names were
John (from Hebrew), William (Germanic), James (Hebrew), George (Greek), Charles
(Germanic), Thomas (Aramaic), Joseph (Hebrew), Henry (Germanic), Samuel
(Hebrew), and Robert (Germanic). Of the 100 most popular male names, 34 were of
Hebrew origin, 22 were Germanic, 16 were English, 11 were Latin, 9 were Greek, 4
were French, 2 were Gaelic, 1 was Aramaic, and 1 was Phoenician. No pronunciation
guide was needed since all the names were familiar.
Phonotactic descriptions
The nineteenth-century name containing the most phonemes was Alexander, with
ten. The names containing the least were Roy and Earl, with two each. Milton, Paul,
and Solomon were the most like Standard American English, with average phoneme
probabilities of .0802, .0729, and .0716. Asa, Hugh, and Isaac were the least, with
average phoneme probabilities of .0128, .0180, and 0189. When bifones were consid-
ered, the names that were most like the English in the corpus were Carl and Warren,
with average bifone probabilities of .0113 and .0108. The names that were the least
like the English in the corpus were Hugh and Earl, both with an average bifone
probability of .0000, and also Roy with .0001. Overall, longer names were more like
English in the Brown Corpus.
Results of phonotactic comparisons
This study made a phonemic comparison of 55 male names from the fictional
Spalding manuscript, 197 male names from The Lord of the Rings, and 100 male
names from the nineteenth-century census records. A one-way ANOVA was used to
compare the three groups of names on their average word length. The nineteenth-
century census names (an average of 5.01 phonemes per name) were on average
shorter than the Spalding names (5.91 phonemes), which were shorter than Lord of
the Rings names (5.97 phonemes). Although the results were statistically significant
(F[2, 349] = 15.84, p < .0001), the effect was not strong (R2 = .083).
Statistically significant differences were also found among the three sources in
average phoneme probability (F[2, 349] = 5.57, p = .0042) compared to the Brown
Corpus, but again the effects were not strong (R2 = .031). The average probability
values for phonemes for names from the Spalding manuscript, Lord of the Rings, and
nineteenth-century census names were .2887, .2591, and .2344 respectively. Differ-
ences in bifone probabilities across the three sources were not significant. A more
detailed analysis comparing the three sources in their patterns of probabilities across
the successive phoneme ordinal positions was needed.
An analysis of pattern was completed to examine how the three name sources
compared at each ordinal position. In particular, we tested four hypotheses:

108 BRAD WILCOX et al.
1. The overall variance of phonemic probabilities will distinguish between natural
naming systems and fictional ones.
2. Fictional naming systems will have more consistency than natural naming
systems in the mean phonotactic probabilities of names of varying word
lengths.
3. Natural and fictional naming systems will differ in the multivariate patterns of
mean phonotactic probabilities at each ordinal position.
4. Distributional properties will distinguish between natural and fictional
systems, with natural being more Gaussian.
First hypothesis
The first hypothesis held that natural naming practices would show greater variance
in all phonemic probabilities and also in all bifone probabilities than would fictional
name systems, with the rationale that names created by single authors would be
expected to be more similar in their phonotactic probabilities than names developing
from a variety of origins within a natural language population. US Census names
represented a natural name distribution. The Spalding manuscript was a fictional
name system. Although Lord of the Rings was also fictional, the author’s competence
with ancient languages led us to expect more sophisticated name creation processes.
The natural naming system showed the greatest variance in phoneme probabilities,
followed by Lord of the Rings, and finally the Spalding manuscript. The 100 names
from the nineteenth century had a variance of phonemic probabilities (combining
across the phoneme ordinal positions) of .01189. In contrast, the 197 names from
Lord of the Rings had a variance of .01030, and the 55 names from the Spalding
manuscript had a variance of .00299. The F ratio used to test for statistical signifi-
cance for the comparison of nineteenth-century names to Lord of the Rings names
was not significant. The comparison of nineteenth-century names to the Spalding
manuscript names was significant (F[99, 54] = 3.97, p = .00000009), and the com-
parison of Lord of the Rings names to Spalding manuscript names was as well (F[196,
54] = 3.45, p = .0000003).
The pattern found in an analysis of the variances of bifone probabilities differed
somewhat in that the nineteenth-century male names and names from Lord of the
Rings were significantly different from one another (F[99, 196] = 1.39, p = .0335).
The variances of probabilities for Spalding manuscript bifones (.00006) was about
one-third as large as that for Lord of the Rings (.00018) and half as large as that for
nineteenth century (.00013). The F ratio comparison of nineteenth-century names to
the Spalding manuscript names was significant (F[99, 54] = 2.06, p = .0021), as was
the comparison of Lord of the Rings names to Spalding manuscript names (F[196, 54]
= 2.87, p = .000008).
Phonotactic probabilities of both phonemes and bifones (each combined across
ordinal position) were highly successful in differentiating between the Spalding man-
uscript fictional names and the nineteenth-century male natural names; however, only
the bifones differentiated between Tolkien’s fictional names and the natural names.
The relationship was not strong, and the variance for Lord of the Rings was slightly
larger than for nineteenth-century names. We determined that more detailed analyses
of these patterns of probabilities across ordinal positions were needed. We compared
the three name sources at each ordinal position individually.

109
IDENTIFYING AUTHORS BY PHONOPRINTS IN THEIR CHARACTERS’ NAMES
Since the Spalding manuscript had no names with fewer than four phonemes or
more than eight, we restricted ourselves to this range and applied the telescoping
method reported for the English vs. Latin pronunciation analysis of this manuscript
to compare names of different lengths. Results are reported here for only the analysis
of four or more phonemes, since the other analyses gave similar results. The analysis
of four ordinal positions included 323 total male names: 55 from the Spalding manu-
script, 179 from The Lord of the Rings, and 89 from the nineteenth-century census
records.
Figure 2 shows the variances of phonotactic probabilities for each of the three name
sources at each of four ordinal positions. The nineteenth-century names and Lord of
the Rings names were both fairly similar, but the Spalding manuscript had signifi-
cantly lower variances in phonotactic probabilities at ordinal positions one, two, and
three, but not four. Because the variances were highest for the nineteenth-century
names, they became the numerator for testing variance ratios for statistical signifi-
cance at each ordinal position. The Spalding variances of phonotactic probabilities
differed significantly from those of the nineteenth century in the first three ordinal
positions (F[88, 54] = 1.58, p = .0355; F[88, 54] = 2.20, p = .0011; F[88, 54] = 1.74,
p = .0149; F[88, 54] = 0.94, n.s.). However, those for Lord of the Rings did not differ
appreciably from those of the nineteenth century (F[88, 178] = 1.24, n.s.; F[88, 178]
= 0.94, n.s.; F[88, 178] = 1.00, n.s.; F[88, 178] = 1.44, p = .0206).
Hypothesis 1 was also tested for the variances of phonotactic probabilities of
bifones. The results are shown in Figure 3. As with the phonemes, the nineteenth-
century and Lord of the Rings names were very similar and the Spalding manuscript
varied substantially. In the first bifone position, the Spalding manuscript names
differed significantly from nineteenth-century names (F[88, 54] = 2.18, p = .0013), and
from Lord of the Rings names (F[178, 54] = 2.12, p = .0009). In the second bifone
position the names from Lord of the Rings and nineteenth century did not differ from
each other, but both had a much higher variance than Spalding (F[178, 54] = 2.87,
p = .000009; F[88, 54] = 2.76, p = .00005). On Bifone 3, none of the variances differed
significantly.
figure 2 The variances of English phonemic probabilities for each of the three sources at
each of the four ordinal phoneme positions.

110 BRAD WILCOX et al.
Hypothesis 1 held that the variance of phonemic probabilities would distinguish
between natural naming systems and fictional ones. While all of the tests differenti-
ated clearly between the Spalding names and the natural nineteenth-century names,
the fictional names crafted by Tolkien were similar to natural naming patterns.
Second hypothesis
The second hypothesis was extended from the first variance hypothesis, but was more
subtle, dealing with variances of mean phonotactic probabilities for names at various
word lengths. We assumed that an author’s artificial naming system would show the
same processes regardless of name length, but that names chosen by people from
varying origins and backgrounds would involve a more heterogeneous set of phono-
tactic structures. The phonotactic probabilities of individual phonemes in natural
naming practice would therefore be expected to vary more across name lengths.
This was tested with a two-way multivariate analysis of variance of the interactive
effects of name source and name length on the ordinal position profiles of phonotac-
tic probabilities. The results from four multivariate tests (Wilks’ lambda, Pillai’s
trace, the Hotelling-Lawley statistic, and Roy’s greatest root) are shown in Table 1
for each of the three sources of variance (name source, name length, and interaction
between these two). All of these 12 multivariate tests were statistically significant,
indicating that phonotactic ordinal position profiles were predictable from length,
source, and the interaction of the two. Name source was a moderating variable, while
name length showed differing effects for each source.
A similar two-way MANOVA was run on the bifone data. The multivariate tests
for name source were not statistically significant. The effects of name length were
significant (Wilks’ lambda [4, 308] = 0.92693, p = .0248), as were the interaction
effects (Wilks’ lambda [8, 308] = 0.84942, p = .0012). The pattern of results from
these parallel analyses on bifones gave similar results and will therefore not be
reported here.
Figure 4 shows the simple effects of name length (from a length of four to a length
of eight) on profiles of phonotactic probability over four ordinal positions. The
figure 3 The variances of English bifone probabilities for each of the three sources at each
of the four ordinal bifone positions.

111
IDENTIFYING AUTHORS BY PHONOPRINTS IN THEIR CHARACTERS’ NAMES
profile at the top, for names of eight phonemes, had the overall highest English prob-
abilities. The profile for names of length four appeared at the bottom, with profiles
for name lengths of five, six, and seven generally lining up in order between these
two. The figure also shows that the second phonemic position had higher English
probabilities than the other three positions.
The univariate tests shown in Table 2 followed-up on the holistic multivariate
results given in Table 1, breaking down the overall statistical significance of profile
comparisons into which particular ordinal positions most accounted for significance.
All three sets of multivariate tests showed statistical significance for the overall gestalt
forms, the holistic patterns, and ten out of the twelve corresponding univariate tests
TABLE 1
THE FOUR MULTIVARIATE STATISTICS FOR A TWO-WAY MANOVA OF PHONEME PROBABILITIES AS A
FUNCTION OF SOURCE OF THE NAMES AND NAME LENGTH
variance source test value F num df den df
p
namesource Wilks’ .9428 2.28 8 610 .0209
Pillai’s .0580 2.28 8 612 .0206
H-L .0598 2.28 8 433.39 .0215
Roy’s .0380 2.90 4 306 .0221
length Wilks’ .8875 2.32 16 932.43 .0023
Pillai’s .1151 2.28 16 1232 .0027
H-L .1238 2.35 16 604.03 .0021
Roy’s .0947 7.29 4 308 <.0001
NxL interaction Wilks’ .8368 1.74 32 1126.4 .0067
Pillai’s .1708 1.72 32 1232 .008
H-L .1861 1.77 32 787.07 .0059
Roy’s .1231 4.74 8 308 <.0001
figure 4 A comparison of the English phonemic probabilities for each word length at each
ordinal position for all three name sources combined.

112 BRAD WILCOX et al.
TABLE 2
FOUR TWO-WAY ANOVAS FOR A TWO-WAY MANOVA OF PHONEME PROBABILITIES AS A
FUNCTION OF SOURCE OF THE NAMES AND NAME LENGTH (ONE ANOVA FOR EACH OF
THE FOUR PHONEME POSITIONS)
ordinal position variance source df SS (Type 3) F
pR2
Phoneme 1 namesource (2, 308) .0051 4.20 .0158 .0247
length (4, 308) .0028 1.15 .3348 .0135
NxL interaction (8, 308) .0135 2.80 .0053 .0657
entire model (14, 308) .0196 2.32 .0048 .0952
Phoneme 2 namesource (2, 308) .0033 2.49 .0844 .0139
length (4, 308) .0051 1.93 .1055 .0216
NxL interaction (8, 308) .0068 1.28 .2523 .0287
entire model (14, 308) .0331 3.56 <.0001 .1394
Phoneme 3 namesource (2, 308) .0012 0.80 .4524 .0050
length (4, 308) .0012 0.38 .8246 .0047
NxL interaction (8, 308) .0064 1.02 .4200 .0255
entire model (14, 308) .0099 0.90 n.s. .0395
Phoneme 4 namesource (2, 308) .0024 2.25 .1068 .0127
length (4, 308) .0107 4.96 .0007 .0566
NxL interaction (8, 308) .0092 2.14 .0317 .0483
entire model (14, 308) .0252 3.34 <.0001 .1319
were also significant. Graphical summaries of the patterns of means were needed to
understand the basis of the significant results.
Figures 5 and 6 show the phonotactic probability profiles as a function of name
length for nineteenth-century names and for Lord of the Rings names. Both of these
followed a similar pattern peaking in the English probabilities at phoneme two, but
the nineteenth-century names had substantially more phonotactic variation than did
the names from Lord of the Rings, particularly in the first ordinal position and in
the fourth. The variance of the mean phonotactic probabilities in the first ordinal
position for nineteenth-century names was 0.288 compared to only 0.018 for Lord of
the Rings (F[4, 4] = 15.77, p = .0044). For the fourth ordinal position the variance
of nineteenth-century names was 0.222 compared to only 0.011 for Lord of the Rings
(F[4, 4] = 21.13, p = .0022).
Figure 7 shows the phonotactic probability profiles for the Spalding names. English
probabilities peaked at phoneme two like the others, but the range of probabilities
at phoneme positions one and four was more comparable to those of nineteenth-
century names than Lord of the Rings names. The variance of the mean phonotactic
probabilities in the first ordinal position for the Spalding manuscript was 0.255 com-
pared to 0.288 for nineteenth-century names (F[4, 4] = 1.13, n.s.). For the fourth
ordinal position the variance of Spalding manuscript names was 0.207 compared
to 0.222 for nineteenth-century names (F[4, 4] = 1.07, n.s.). Although the Spalding
manuscript names were comparable to the nineteenth-century names in variances of

113
IDENTIFYING AUTHORS BY PHONOPRINTS IN THEIR CHARACTERS’ NAMES
mean probabilities, the patterns of the probability profiles shown in Figure 7 were
much less orderly than the patterns in Figures 5 and 6, which showed a clear trend
for longer names to be more English like.
Figure 8 shows the variances of mean phonotactic probabilities for each of the three
sources at each phonemic ordinal position. The pattern of variances for Spalding
figure 5 A comparison of the English phonemic probabilities for each word length at each
ordinal position, names from the 19th century.
figure 6 A comparison of the English phonemic probabilities for each word length at each
ordinal position, names from Lord of the Rings.

114 BRAD WILCOX et al.
figure 8 A comparison of the variances across the five word lengths of names for each of
the three sources, at each of the first four ordinal positions.
names and nineteenth-century names was U-shaped, whereas the pattern for Lord of
the Rings names was inverted.
The tests of the first hypothesis showed a clear separation between the natural
names and the Spalding names, but not Lord of the Rings names. The tests of the
second hypothesis showed clear evidence of the separation of the natural names from
the Lord of the Rings names but not from the Spalding names.
Third hypothesis
Whereas the first hypothesis explained the variances of phonotactic probabilities and
the second hypothesis was concerned with the variances of means within each ordinal
figure 7 A comparison of the English phonemic probabilities for each word length at each
ordinal position, names from the Spalding manuscript.

115
IDENTIFYING AUTHORS BY PHONOPRINTS IN THEIR CHARACTERS’ NAMES
TABLE 3
THE FOUR MULTIVARIATE STATISTICS FOR A ONE-WAY MANOVA OF PHONEME PROBABILITIES AS A
FUNCTION OF NAME SOURCE (RESULTS FOR ENTIRE MODEL AND FOR TWO LINEAR CONTRASTS)
Variance Source Test Value
F
Num df Den df
p Multivariate R2
Entire Model Wilks’ .8891 4.80 8 634 <.0001 .1109
Pillai’s .1137 4.79 8 636 <.0001
H-L .1216 4.81 8 450.54 <.0001
Roy’s .0843 6.70 4 318 <.0001
Contrast A:
Natural vs. Fictional
Wilks’ .9225 6.66 4 317 <.0001 .0775
Pillai’s .0775 6.66 4 317 <.0001
H-L .0840 6.66 4 317 <.0001
Roy’s .0840 6.66 4 317 <.0001
Contrast B:
LotR vs. Spalding
Wilks’ .9615 3.17 4 317 .0142 .0385
Pillai’s .0385 3.17 4 317 .0142
H-L .0400 3.17 4 317 .0142
Roy’s .0400 3.17 4 317 .0142
position, the third hypothesis deals with the patterning of the mean probabilities
themselves. It held that natural and fictional naming systems would differ in the
multivariate patterns of mean phonotactic probabilities across the ordinal positions.
We employed a one-way multivariate analysis of variance (MANOVA with contrasts)
to test the pattern match of means across the three sources, with source as the
independent variable and phonotactic probabilities at each ordinal position as the
dependent variable. Although we conducted a similar one-way MANOVA analysis
with bifone ordinal positions as the dependent variable, none of the statistical tests
reached significance. Therefore, we report the results only for phonemes.
Tables 3 and 4 show the results of the one-way MANOVA and its associated
ANOVAs for phonemes.1 Both tables report three analyses: the significance level for
the entire model comparing the three sources, the significance level for Contrast A of
the two fictional naming systems compared to nineteenth-century names, and the
significance level for Contrast B of Lord of the Rings compared to Spalding.
The results from four multivariate tests are shown in Table 1. The three sets of
multivariate tests were each statistically significant.
In Contrast A of Table 1, natural naming systems (nineteenth century) were
differentiated from fictional naming systems. The profile for the unsophisticated
Spalding names differed significantly from that for the relatively sophisticated system
in Lord of the Rings.
No significant model differences were found at the third phoneme position, but the
entire model was significant in all other phoneme positions. Phoneme two was the
only position in which all three tests were significant, as evident in the pattern shown
in Figure 9, where phoneme two was the point where the three curves were most
differentiated. The mean phonemic probability for Spalding (.070) was higher on
phoneme two than that of Lord of the Rings (.060) which was higher than nineteenth-
century names (.051).

116 BRAD WILCOX et al.
figure 9 A comparison of the three name sources in their average English phonemic prob-
abilities at each of the four ordinal positions.
Names from fictional sources were significantly different than natural names from
the nineteenth century (Contrast A) in the second and fourth phoneme positions.
Names from Spalding were significantly different from names from Lord of the Rings
(Contrast B) in the first and second phoneme positions. Overall, Spalding names were
most English-like and nineteenth-century names were least English-like.
Hypothesis 3 held that natural and fictional naming systems would be differenti-
ated by the multivariate patterns of mean phonemic probabilities. Table 3 shows
strong and significant effects for Contrast A, natural versus fictional, accounting for
about 8% of the variance.
TABLE 4
FOUR ONE-WAY ANOVAS SHOWING RESULTS FOR THE ENTIRE MODEL AND FOR THE TWO LINEAR
CONTRASTS (ONE ANOVA FOR EACH OF THE FOUR PHONEME POSITIONS)
Ordinal Position Variance Source df SS
FpR2
Phoneme 1 Contrast A: Natural vs. Fictional (1, 320) .0012 1.92 .1664 .0059
Contrast B: LotR vs. Spalding (1, 320) .0036 5.67 .0178 .0174
Entire Model (2, 320) .0039 3.07 .0480 .0188
Phoneme 2 Contrast A: Natural vs. Fictional (1, 320) .0111 15.68 <.0001 .0465
Contrast B: LotR vs. Spalding (1, 320) .0038 5.46 .0201 .0162
Entire Model (2, 320) .0120 8.51 0.0002 .0505
Phoneme 3 Contrast A: Natural vs. Fictional (1, 320) 0.0001 0.07 .7918 0.0002
Contrast B: LotR vs. Spalding (1, 320) 0.0003 0.36 .5469 .0011
Entire Model (2, 320) 0.0003 0.19 .8312 .0012
Phoneme 4 Contrast A: Natural vs. Fictional (1, 320) .0056 9.99 .0017 .0296
Contrast B: LotR vs. Spalding (1, 320) .0015 2.64 .1051 .0078
Entire Model (2, 320) .0099 8.77 0.0002 .0520

117
IDENTIFYING AUTHORS BY PHONOPRINTS IN THEIR CHARACTERS’ NAMES
Fourth hypothesis
The fourth hypothesis held that distributional properties could also distinguish
between historical naming systems and individual author systems. Tests for normal-
ity (Gaussian shape of the data distributions) were of particular interest as descriptive
measures for differentiating the three sources.
The twelve name distributions were tested for fit to a Gaussian curve using the
Shapiro-Wilk test of normality (Shapiro and Wilk, 1965), and the Shapiro-Francia test
of normality (Shapiro and Francia, 1972). They were also tested graphically for fit
using the Q-Q plot. In the second phoneme position, the natural nineteenth-century
names had a relatively Gaussian distribution, both in the numerical tests and in the
graphical Q-Q plot — not surprising, as the second phoneme position is often a
vowel in the center of the first morpheme within a name. The other three phoneme
positions for nineteenth-century names had statistically significant departures from
Gaussian shape. For Lord of the Rings all four phoneme positions departed signifi-
cantly from normality, and for the Spalding manuscript all did but phoneme position
three.
Figure 10 compares the Q-Q plots for each of the three sources in the second
ordinal position. The Y axis of this plot shows the quantile values from the distribu-
tion in question, and the X axis shows the corresponding quantile values for a normal
distribution. When the data are normally distributed, the plotted bivariate points
form a straight diagonal line from the bottom left of the figure to the upper right.
Superimposed upon the Q-Q plots for each name source are the results of the tests
of normality.
The repetition of one particular phoneme in Spalding names, and two in Lord
of the Rings are obvious. These were a large part of the basis for the departure in
normality in the two distributions. For Spalding, the particularly common phoneme
was /æ/ as in cat or black; 30 of the 55 names in the Spalding manuscript (54.6%)
used this phoneme in the second position (e.g., Bambo, Baska, Fabious, Gamasko,
Habelan, Hamack, Dato, Lambon, Rambock, Sambol). Tolkien’s two most common
phonemes in the second ordinal position were /r/ as in red and try, and /oʊ/ as in go
and home; 42 of the 179 names in Lord of the Rings (23.5%) used /r/ in the second
position (e.g., Aragorn, Brandir, Bregor, Frodo, Grishnak), and 21 of the 179 (11.7%)
used /oʊ/ in the second position (e.g., Boromir, Lorien, Nori, Roac). Repetition
patterns of this type were not so obvious in the more Gaussian pattern for the
nineteenth-century names.
These differences between the natural naming pattern of the nineteenth century and
the patterns for the two fictional systems are brought into perspective by comparing
the percentage of names accounted for by the two most common phonemes in the
second ordinal position and also in the other three positions as shown in Figure 11.
In both fictional name systems, the percentage of names accounted for by the two
most common phonemes became less for each successive ordinal position from the
second phoneme on, whereas for the nineteenth-century names the percentage
increased with each successive ordinal position. There was a statistically significant
difference in profile trends of the Figure 11 percentages between natural names and
both of the fictional naming systems. For Spalding names compared to nineteenth-
century names the chi square value was found to be large (χ2(3) = 24.417, p < .0001).

118 BRAD WILCOX et al.
figure 11 Bar graphs showing the percents of names from each of the three sources
accounted for by the two most common phonemes in each of the four ordinal positions.
figure 10 Q-Q plots for examining departure from Gaussian distributional shape: A com-
parison of the three manuscript sources at the second phoneme position. Also shown are the
p values for the Shapiro-Wilk and the Shapiro-Francia tests of normality.

119
IDENTIFYING AUTHORS BY PHONOPRINTS IN THEIR CHARACTERS’ NAMES
It was somewhat smaller for Lord of the Rings names compared to nineteenth-
century names, but still significant (χ2(3) = 11.360, p < .0099). The profile trend of
Lord of the Rings names did not differ significantly from that for Spalding names
(χ2(3) = 0.923, p < .8198), indicating clear differentiation between natural and
fictional naming patterns.
Conclusion
The four hypotheses were selected to test whether natural naming systems could be
distinguished from fictional ones, first by overall variance of phonemic probabilities,
second by the variance in mean probabilities across word lengths at each ordinal
position, third by multivariate profiles of mean phonotactic probabilities, and fourth
by the distributional properties of the phonemic probabilities. The test of the first
hypothesis distinguished Spalding names from natural names, but Lord of the Rings
names were not distinguished from natural. In contrast, the test of the second hypoth-
esis distinguished Lord of the Rings names from natural, but not Spalding names.
However, on both Hypothesis 3 and Hypothesis 4, both of the fictional sources of
names were clearly distinguished from natural names.
This research was an exploratory study to determine whether there were sufficient
phonemic differences between fictional and authentic names to merit further investi-
gation. Results indicated that there may be a phonoprint of sorts in the fictional work
of both Tolkien and Spalding that may also surface in the work of other authors.
Many more fictional and natural names from a variety of authors and time periods
will need to be analyzed before final conclusions can be drawn.
Results of this study indicated that while it is possible to create a convincing set of
names for a story, as Spalding and Tolkien did, such names seemed to follow patterns
at the phoneme and bifone levels that were significantly different than those of
authentic lists of names from a variety of cultural origins. The possibility of phono-
prints invites further investigation, and the methods of analysis used in this study may
lead to new ways of doing so.
Note
1 Both ANOVA and MANOVA assume a population
that is normally distributed (Gaussian). However,
like the t test from which it is derived, ANOVA is
robust with respect to violation of this assumption
(Box, 1953), particularly when sample size is large,
as it was here. Note in the discussion of Hypothesis
4 in the next section, however, that Gaussian shape
of phoneme probabilities might be useful as a dif-
ferentiator between natural naming systems and
fictional ones.
Bibliography
Archer, John B., John L. Hilton, and G. Bruce Schaalje. 1997. “Comparative Power of Three Author-Attribution
Techniques for Differentiating Authors.” Journal of Book of Mormon Studies 6(1): 47–63.
Bailey, Todd M. and Ulrike Hahn. 2001. “Determinants of Wordlikeness: Phonotactics or Lexical Neighbor-
hoods?” Journal of Memory and Language 44(4): 568–591.
Baily, Richard W. 1979. “Authorship Attribution in a Forensic Setting.” Advances in Computer-Aided Literary
and Linguistic Research: 9.
Box, George Edward Pelham. 1953. “Non-Normality and Test on Variance.” Biometrika 40: 318–335.

120 BRAD WILCOX et al.
Chance, Jane. 2004. Tolkien and the Invention of Myth: A Reader. Frankfort, KY: University of Press of
Kentucky.
Croft, D. James. 1981. “Book of Mormon ‘Wordprints’ Reexamined.” Sunstone 6(2): 15–22.
Dell, Gary S., Kristopher D. Reed, David R. Adams, and Antje S. Meyer. 2000. “Speech Errors, Phonotactic
Constraints, and Implicit Learning: A Study of Experience in Language Production.” Journal of Experimental
Psychology: Learning, Memory, & Cognition 26(6): 1355–1367.
Downey, Sean S., Brian Hallmark, Murray P. Cox, Peter Norquest, and J. Stephen Lansing. 2008. “Computa-
tional Feature-Sensitive Reconstruction of Language Relationships: Developing the ALINE Distance for
Comparative Historical Linguistic Reconstruction.” Journal of Quantitative Linguistics 15(4): 340–369.
Eldredge, J. Lloyd. 2005. Teach Decoding: Why and How (2nd ed.). Upper Saddle River, NJ: Pearson
Education.
Erickson, Whitney, Angel L. Venegas, Alexandra Brattos, Malvania Salash, Donald W. Walker, Jessica H. Scott,
and Bruce L. Brown. 2008. “Convergent Multivariate Graphics for Capturing the Naming Patterns in US
Census Data from the Nineteenth Century.” Proceedings of the Section on Government Statistics. Alexandria,
VA: American Statistical Association.
Francis, W. Nelson. 1981. “Word-Making: Some Sources of New Words.” Language: Introductory Readings. Ed.
Virginia P. Clark, Paul A. Sschholz, and Alfred F. Rosa. New York: St Martin’s Press, 316–328.
Hanks, Patrick and Flavia Hodges. 1995. A Dictionary of First Names. New York: NY: Oxford University
Press.
Hilton, John L. 1990. “On Verifying Wordprint Studies: Book of Mormon Authorship.” BYU Studies 30(3):
89–108.
Holmes, David I. 1994. “Authorship Attribution.” Computers and the Humanities 28(2): 87–106.
Iqbal, Farkhund, Liaquat A. Khan, Benjamin C. M. Fung, and Mourad Debbabu. 2010. “E-mail Verification of
Forensic Investigation.” Available at: <http://users.encs.concordia.ca/~fung/pub/IKFD10sac.ppt> (Accessed 23
March, 2013).
Jusczyk, Peter W., Angela D. Friederici, Jeanine M. I. Wessels, Vigdis Y. Svenkerud, and Ann Marie Jusczyk.
1993 “Infants’ Sensitivity to the Sound Patterns of Native Language Words.” Journal of Memory & Language
33: 402–420.
Jusczyk, Peter W., Paul A. Luce, and Jan Charles-Luce. 1994. “Infants’ Sensitivity to Phonotactic Patterns in the
Native Language.” Journal of Memory & Language 33(5): 630–645.
Kessler, Brett and Rebeca Treiman. 1997. “Syllable Structure and the Distribution of Phonemes in English
Syllables.” Journal of Memory & Language 37(3): 295–311.
Kucera, Henry and Nelson W. Francis. 1967. Computational Analysis of Present-day American English.
Providence, RI: Brown University Press.
Morton, Andrew Q. 1979. Literary Detection. New York, NY: Scribner’s Sons.
Rencher. A. C. 2002. Methods of Multivariate Analysis (2nd ed.). New York: John Wiley and Sons.
Shapiro, Samuel and R. S. Francia. 1972. “An Approximate Analysis of Variance Test for Normality.” Journal of
the American Statistical Association 67: 215–216.
Shapiro, Samuel and M. B. Wilk. 1965. “An Analysis of Variance Test for Normality (Complete Samples).”
Biometrika 52(3/4): 591–611.
Shih, Stephanie. 2012. “Linguistic Determinants of English Personal Name Choice.” Paper presented at the
annual conference of the Linguistic Society of America, Portland Oregon.
Starr, Rebecca. 2012. “Disambiguating Romanized Chinese Personal Names: A Corpus Based Approach to
Backtransliteration and Gender Identification.” Paper presented at the annual conference of the American
Name Society, Portland, OR.
Storkel, Holly L. 2001. “Learning New Words: Phonotactic Probability in Language Development.” Journal of
Speech, Language and Hearing Research 44: 1321–1337.
Storkel, Holly L. 2003. “Learning New Words II: Phonotactic Probability in Verb Learning.” Journal of Speech,
Language, & Hearing Research 46: 1312–1323.
Vitevich, Michael S. 2002. “Influence of Onset Density on Spoken-word Recognition.” Journal of Experimental
Psychology: Human Perception and Performance 28: 270–278.

121
IDENTIFYING AUTHORS BY PHONOPRINTS IN THEIR CHARACTERS’ NAMES
Vitevich, Michael S., David B. Pisoni, Karen I. Kirk, Marcia Hay-McCutcheon, and Stacey L. Yount. 2002.
“Effects of Phonotactic Probabilities on the Processing of Spoken Words by Postlingually Deafened Adults with
Cochlear Implants.” Volta Review 102: 283–302.
Vitevich, Michael S. and Paul A. Luce. 2004. “A Web-based Interface to Calculate Phonotactic Probability for
Words and Nonwords in English.” Behavior Research Methods, Instruments, & Computers 36(3): 48–487.
Vitevich, Michael S. and Paul A. Luce. 1999. “Probabilistic Phonotactics and Spoken-word Recognition.” Journal
of Memory & Language 40(3): 374–408.
Vitevich, Michael S., Paul A. Luce, Jan Charles-Luce, and David Kemmerer. 1997. “Phonotactics and Syllable
Stress: Implications for the Processing of Spoken Nonsense Words.” Language and Speech 40: 47–62.
Whissell, Cynthia. 2001. “Sound and Emotion in Given Names.” Names 49(2): 97–120.
Wright, Saundra K. 2012. “Naming Decisions Made by International Students Studying in the US”. Paper
presented at the annual conference of the American Name Society, Portland, OR.
Wu, Rui-Wen. 2010. “Development and Strata Analysis of Geng She Unrounded Cognates in Proto-Min.”
Language and Linguistics 11(2): 297–334.
Young, Steven. 2004. “‘Old Prussian’ in M. Pratorius’s Deliciae Prussicae.” Studies in Baltic and Indo-European
Linguistics: In Honor of William R. Schmalsteig. Ed. Philip Baldi and Pietro U. Dini. Amsterdam, Netherlands:
John Benjamin, 275–283.
Zheng, Rong, Yi Quin, Zan Huang and Hsinchun Chen. 2003. “Authorship Analysis in Cybercrime Investiga-
tion.” Available at: <http://www.mendeley.com/catalog/authorship-analysis-cybercrime-investigation/> (Accessed
24 March, 2013).
Notes on contributors
Brad Wilcox is an Associate Professor in the Department of Teacher Education at
Brigham Young University where he teaches graduate and undergraduate courses
in literacy and children’s literature. His research interests include reading, writing,
education in international settings, and onomastics.
Correspondence to: Brad Wilcox, 201-P MCKB, Brigham Young University, Provo,
UT 84602, USA. Email: brad_wilcox@byu.edu
Bruce L. Brown is a Professor in the Department of Psychology at Brigham Young
University where he teaches graduate and undergraduate courses. His research
interests include psychology, statistics, and the study of names.
Wendy Baker-Smemoe is an Associate Professor in the Department of Linguistics and
English Language at Brigham Young University. Her research interests include second
language acquisition, dialectology, and onomastics.
Sharon Black is an Associate Teaching Professor and editor/writing consultant in the
David O. McKay School of Education at Brigham Young University. Her research
interests include early literacy instruction, gifted/talented education, and elementary
arts integration.
Justin Bray graduated from Brigham Young University in 2011 with a BA in history
and a minor in Latin American studies. He currently works at the LDS Church
History Library in Salt Lake City, Utah.