Working within three dimensions is extremely advantageous from a Communication
communications point of view, but in many practical instances this is rarely through cartoon
possible if more than a superficial overview is to be conveyed. Many alternative
pictorial methods have been employed in an attempt to facilitate the
communication of information – ranging from the familiar bar and pie charts
and pictograms to more obscure forms. The pie chart, bar chart and trend
graph as detailed by Beattie and Jones (1992), have become familiar and 69
acceptable in the financial report as alternatives to the narrative and numerical
form; schematic faces have yet to achieve such acceptability, perhaps because of
the novelty of the approach and the emotive manner in which their accounting
message is conveyed.
Alternative methods of presentation, notably those involving the use of a
facial format, may seem a little strange to existing users, but the test of their
usefulness will be in the successful communication of financial messages. While
demanding attention, such figures should also be clear enough to make
interpretation possible without a detailed explanation. The complexity and
familiarity of faces makes them a special class of visual input which derives
from developmental changes in infants, whereby they learn quickly to respond
to more differentiated forms. Schaffer (1971, p. 69), suggests that with
increasing age the overriding importance of the eyes as a source of recognition
and attraction is complemented by increasing attention to other facial features,
facilitating the differentiation between various expressions. The similar
reaction of infants to real faces, photographic representations and schematic
line drawings, forms the basis of their reaction as adults to the messages
provided by cartoon faces. The possibility exists that, with appropriate
assignments, the facial format might be employed to communicate information
on the magnitude and change in a number of variables simultaneously without
the need for detailed explanation or education of users.
There has been limited study to date of the effectiveness of alternative
methods of presenting accounting information for financial decision purposes.
Smith and Taffler (1984) recommend the further exploration of the use of
schematic faces to represent accounting information, following the success of
this medium in displaying multivariate data in other task environments. This
paper explores empirically the usefulness of the schematic face as a
communication device, in a particular decision context, compared with more
conventional presentation formats, focusing on the relative usefulness of
schematic faces, financial ratios and accounting statements as information
formats for decision making.
Financial information is both complex and multidimensional and if a complete
picture is to emerge, rather than a series of financial relationships, then
additional graphical methods are required which will represent adequately the
multivariate nature of financial data. Canadian Institute of Chartered
Accountants CICA (1993, p. 122) suggests that the ability of multivariate
AAAJ graphics to portray data in an understandable form may result in their
9,2 producing better decision making than corresponding tabular presentations.
The impact of graphical representation is an important attribute since a
forceful picture must be produced which allows new stimuli from a complex
data set to be perceived while existing stimuli are being integrated.
Psychologists, among them Yin (1969), Smith and Nielsen (1970), and Reed
70 (1972), argue that the acquisition and organization of information within
dimensions, by decision makers, is perceived as a Gestalt so that stimuli are
processed in a holistic manner. In addition to providing a Gestalt, the familiarity
of faces commands attention and triggers an emotional reaction which
enhances their overall impact in a way that other forms of graph do not.
Garner (1978), Homa et al. (1976) and Sergent (1984) provide empirical
support for the face being regarded as a spatial interrelationship of features
capable of being perceived as a Gestalt, an issue central to the development of
the schematic face in cognitive research. The familiarity of faces and their ease
of recognition and description makes them superior to other pictorial forms of
representation. Wilkinson (1982) demonstrates that face-based icons
outperform alternative formats in the judgement of similarities. Haig (1984)
demonstrates the incredible sensitivity of respondents to the smallest changes
in facial features. Morton and Johnson (1989) note that faces are special more
than by virtue of their being visible parts of the human form since they can
signal their intentions. This is so even though there is no convincing evidence
that they are processed any differently from other objects among which we
require to discriminate.
Chernoff (1971) initiated the computer-based construction of schematic
“faces” whose features can be made to vary in size and shape according to the
value of the assigned variable. The original form of portrait has been
adapted by Bruckner (1978) to provide greater variation and by Frith in Everitt
(1978) and by Flury and Riedwyl (1981) to provide greater realism. Valentine
(1986) views the human face as a series of vectors in multidimensional space
with dimensions corresponding to significant features. Her study suggests that
a matching of significant features with financial performance measures,
provides the possibility of communicating multidimensional financial
information in a simple, integrated and readily comprehensible form.
Three major contributions to the literature in this area (Moriarity, 1979;
Smith and Taffler, 1984; Stock and Watson, 1984) have applied Chernoff ’s
methods in a financial environment: Moriarity (1979), in the seminal study in
the area, working with financial statement data, examines the use of
multidimensional graphics as a technique for describing the financial status of
the firm. His innovative approach provides encouraging results which suggest
that unsupported faces provide an excellent framework for decision making
when produced as an alternative to information conveyed in more traditional
fashion. Moriarity examines the speed and accuracy with which respondents
classify companies as failed or non-failed, without knowledge of prior
probabilities, when presented with appropriate financial information in Communication
alternative formats. through cartoon
Moriarity’s respondents found the changes most easily detected in the faces, graphics
which were classified faster and more accurately than either the raw accounting
numbers or derived ratios. His results suggest that our familiarity with faces
makes an interpretation of the portraits possible without a detailed knowledge
of the information used or the facial assignments employed. However, 71
Moriarity’s sample included only ten respondents of any accounting
sophistication, the remainder being first year accounting undergraduates, so
the strong relative performance with the faces may be attributable to ignorance
of accounting concepts. Moriarity makes no attempt to match the most salient
financial and facial features nor to manipulate eyebrow slant in his schematic
faces, despite the psychological evidence suggesting their importance.
Stock and Watson (1984) take a similar approach but employ judgementally-
determined bond ratings as their classification base in suggesting the potential
usefulness of schematic faces in situations where statistical models are weak.
They demonstrate the relative usefulness of faces but their findings may reflect
the complexity of the task rather than the method of data presentation.
In addition, both of the above studies might be criticized for failing to
compare like with like, so that the superiority of presentation apparent from the
faces may actually represent superiority of information. Schematic faces are
constructed in a relative, not absolute manner, usually standardized relative to
industry means and standard deviations. However, Moriarity provides no
standard deviations for his financial ratios and Stock and Watson provide
neither means nor standard deviations. No attempt is made to compare the
performance of users of varying levels of accounting sophistication.
Smith and Taffler (1984) in a UK environment illustrate how intertemporal
performance comparisons can be made using schematic faces together with
their use in facilitating the distinction between failed and non-failed companies
in large datasets. They suggest that schematic faces may provide a clearer
indication of financial status than is apparent from a company’s financial ratio
profile. However, their study does not show whether accounting data can be
analysed more quickly or more effectively in a facial format than when
represented by more conventional means. None of these studies adequately
reflects contributions from the psychological literature relating to feature
assignment and facial construction. These are addressed below with a
consideration of the features of the face, their interaction, and the use of
Goldstein and Mackenberg (1966), Grant (1970), Laughery et al. (1971) and
De Soete and De Corte (1985) identify three main expressive areas of the face
which link movements with particular emotions: the eyes, the eyebrows and the
mouth. They emphasize the importance of the eye and mouth regions, which are
more mobile than others and communicate more information, and the relative
insignificance of the ears. Notably, Stock and Watson (1984) employ a feature
direction the opposite of that suggested by the psychological literature and
AAAJ accord the nose a prominent role in the assignment of financial variables to
9,2 facial features, despite the strong psychological evidence to the contrary.
McKelvie (1973) demonstrates that the perceived meaningfulness of the face
is at its greatest when both the eyebrows and mouth curvature vary from an
average position, suggesting that the interaction of eyebrow slant and mouth
curvature provides an effective force in the communication of meaning in facial
72 expressions. The findings of these studies from the psychological domain are
employed in the facial constructions in this study.
Following Ryan and Schwartz (1956), Chernoff (1978) suggests that
experience with caricatures and cartoons indicates that the need for realistic
faces on pictures is not great. Further supporting evidence is supplied by
Diamond and Carey (1986), who find respondents to prefer schematic cartoon
faces to real faces in recognition exercises, and by Rhodes et al. (1987) who show
that facial caricatures are recognized more quickly than line drawings – results
consistent with a holistic theory of encoding. Benjamin and Pachella (1982)
advise against making Chernoff faces more realistic, since the consequent
introduction of irrelevant information will cause perceptual problems. They
suggest that respondents will be unable to ignore irrelevant features, even when
instructed to do so, so that the number of features presented in the display
should be equal to the number of variables whose values are to be mapped.
The suggestion is that where the facial portrait is required to communicate a
message the emphasis must be placed on the mobile features. These features
can be varied efficiently with the Chernoff (1978) and Bruckner (1978)
formulation to facilitate the interpretation of the overall portrait, so that
financial performance can be represented through appropriate assignment of
financial attributes to facial characteristics. Provided that due attention is paid
to the combination of facial features, without overemphasis on dominant
features, it is possible that an integrative picture might emerge to give a clear
indication of overall performance.
This study overcomes many of the deficiencies of its predecessors and makes
original contributions by incorporating relevant evidence from the
psychological literature, and by extending the experimental work to skilled
users. The complex issue of how financial variables are assigned to facial
characteristics, however, remains an area for further study. No attempt is made
here to vary the feature assignment, rather a single assignment is employed
throughout; the most salient financial variables are assigned to those facial
features deemed by the literature to be the most important, in a manner entirely
consistent with the psychological evidence.
The task domain is the failed/non-failed company decision situation since there
is a wealth of literature demonstrating the strong degree of environmental
predictability for accounting statement-based ratio information (e.g. Altman,
1968; Taffler, 1982). A substantial literature, summarized by Foster (1986,
p. 534), highlights the prediction of performance on the basis of trends supplied
by financial ratios and accounting statements. The ease of processing facial Communication
profiles suggests that they might provide a more efficient means of making the through cartoon
same analysis. This paper suggests that the relative lack of predictability of graphics
information presented in the form of schematic faces in existing studies may be
due to insufficient attention both to the appropriate assignment of variables to
facial components and the relevant psychological literature on feature saliency.
Accounting ratios suggested by Taffler and Sudarsanam (1980) are used to 73
represent the four key dimensions of accounting information. These ratios
are assigned to facial features in order to provide Chernoff portraits of the type
shown in Figure 1. The aim is to produce faces which reflect the financial
performance of the company and which can be interpreted without the need for
a detailed explanation of the variables employed or the feature assignments.
A template for failure
from the assignment of
financial variables to
Distressed Neutral Healthy
The schematic Chernoff faces employed here incorporate four variable features:
mouth, eyes, eyebrows and nose. The ears remain constant throughout but the
size and shape of the face may change because of the size and position of the
facial features which it bounds. The precise location of each of the features is
determined by the values of the assigned financial ratios relative to industry
means and standard deviations. Thus, mouth (length, curvature and height)
will be determined by a profit ratio; eye (separation, size and pupil direction)
will be determined by financial gearing; eyebrow (angle and height) will be
determined by liquidity, and nose (length and width) will be determined by
working capital position. Different assignments are possible, but this particular
assignment ensures that three key financial performance measures (profit,
gearing and liquidity) are assigned to the mobile features of the face.
Moriarity (1979), Stock and Watson (1984) and Smith and Taffler (1984) all
use this type of schematic face structure to contrast “healthy” and “distressed”
companies. An impression of a healthy, profitable and secure company is
created by a smiling face and large eyes, while a company in financial distress
has a worried frown, down-turned mouth and small eyes. The overall message
created for the latter would be one of the empty, washed-out face of an
AAAJ The clarity with which the foregoing figures convey their financial
9,2 messages, in the absence of a detailed narrative explanation, provides the
impetus for an analysis of the relative explanatory power of alternative
The feature assignment illustrates the way in which a visual impression of
the personality of each company can be created which may even allow some
74 speculation as to its corporate identity and attributes. The financial ratios are
mapped precisely onto the facial features so that their positions correspond
precisely relative to industry averages.
For each financial ratio calculated, the industry means and standard
deviations corresponding to that time period are used to convert the ratio into a
numerical form expressed as “number of standard deviations above/below the
For example, suppose that the manufacturing Company XYZ has a
profitability ratio (PBIT/TA) of 0.238, where the industry mean and standard
deviation for that sector in that particular year are respectively 0.07 and 0.12.
Relative to the rest of the sector Company XYZ’s profitability is, therefore, 1.4
standard deviations above the mean
0.238 – 0.07
The mean position for the neutral face and the total range of lengths and angles
feasible in the facial caricature allow the development of means and standard
deviations for each facial feature. Financial ratios are mapped onto their
assigned facial characteristic in terms of the number of standard deviations
from the mean so that a precise correspondence of number to position is
achieved. If profitability is mapped onto the mouth, say, then both the length
and curvature of the mouth will be determined by the profit ratio. In the case of
Company XYZ, above, both the length and the curvature will be 1.4 standard
deviations above their mean position, resulting in the display of a modest smile.
Use of industry relatives means that it is possible for an improved financial
ratio in absolute terms to coincide with a deterioration of facial message if the
ratio improvement is lower than that for the industry as a whole. The implicit
use of industry statistics in constructing the facial portrait potentially improves
the processing of financial messages over that with financial ratios, even when
the industry statistics are made available to users.
Consequent on the results of earlier studies, together with the incorporation
of advances in the psychological literature, two issues arise for further
(1) facial profiles might be processed significantly more quickly than either
financial ratios or accounting statements;
(2) the classification decisions made using the facial profiles might be more Communication
accurate than those made with either financial ratios or accounting through cartoon
Together, these issues prompt a discussion of the resultant improvement in
decision performance from the use of schematic faces, measured in terms of
“efficiency” and “effectiveness”. With respect to the format of presentation, the
most efficient form of presentation is defined by Bertin (1983) as that which 75
minimizes the effort, measured by time, which is expended to interpret the
relevant aspects of the information set and provide a satisfactory answer to the
given question. Effectiveness is defined by Lusk (1979), in this context, as the
form of presentation which makes it easiest to generate the most accurate
answer to a given set of questions.
An experiment is, therefore, conducted to examine the facility of respondents
of varying accounting sophistication with accounting information presented in
alternative formats. Their processing time and the number of errors of
classification that they make generate measures of the efficiency and
effectiveness of the alternative formats.
Although several previous studies (e.g. De Sanctis, 1984; Remus, 1984) have
compared graphical and tabular data presentation formats, few have used
schematic faces. Mackay and Villarreal (1987) recognize that Chernoff displays
capture multivariate data holistically, in a mnemonic way,unique among
graphical presentations, making comparisons with other presentation forms
potentially difficult. Their study fails to identify any superiority of Chernoff
faces over tabular data in financial decisions, but they do not distinguish
between Type I and Type II errors. Altman et al. (1977) suggest a relative
misclassification cost weighting of 40:1 in favour of Type I errors, relative to
Type II errors, suggesting that a processing format is required which minimizes
opportunities to misclassify failed companies.
While the simplicity of the facial technique is a positive feature in
communicating financial information, especially to the less sophisticated of
users of accounting information, it can be a barrier preventing its widespread
use. It has to be demonstrated that, apart from the novelty of approach, this
method can improve on the quality of decisions made using traditional
methods. To test the hypothesis that facial profiles might provide an efficient
means of representing financial variables, an experiment is devised to test the
reactions of respondents to financial information expressed in alternative
• accounting statements;
• financial ratios derived from these statements; and
• “faces” constructed by the application of financial ratios to particular
The experiment is conducted with three different groups of skilled users and a
group of naïve users to represent users of all levels of sophistication:
AAAJ accounting academics from the Universities of Leeds, Lancaster and
9,2 Birmingham, accounting practitioners from Big 6 companies and MBA finance
majors from City University, London. Together they provide a sample size of
121 sophisticated users, comprising 52 accounting academics, 23 practitioners
and 46 postgraduate students. The skilled users, though not all practitioners,
employ accounting information and financial statement data regularly, and
76 have an extensive knowledge of accounting terminology and format.
Comparisons with naïve users, unfamiliar with accounting information, proved
unworkable at the pilot stage of the study causing them to be excluded from the
sample. These relatively unsophisticated users responded well to the schematic
faces but for them the faces were the only accounting medium to convey any
A systematic sample of 20 listed UK manufacturing companies is chosen to
provide a majority that are still trading and a minority of failures, together with
examples of companies across the whole range of processing difficulty. A
random sample of a company database is not employed since it would be
unlikely to give many (if any) failures. A 14:6 (i.e. 70 per cent:30 per cent) split
between non-failed and failed companies in the sample is adopted because it
conforms closely with the percentage split at the time between healthy and
distressed companies in the population, based on their computed Z-scores. This
division avoids an even distribution of companies while providing enough
variety in the sample to illustrate the performance range. Companies with
financial year ends between 1974 and 1980 are chosen to reflect the clearly
healthy/clearly failed extremes while including several marginal and
potentially more difficult cases. At no time are the respondents made aware of
the 14:6 division. Accounting statements and financial ratios are prepared
and faces constructed for each of the 20 companies over five-year periods and a
random numbering system used to separate the statements/ratios/faces
Respondents are familiarized with the use of schematic faces during a
20-minute briefing session immediately prior to the conduct of the experiment.
The briefing addresses:
• accounting information as a complex multivariate dataset;
• alternative graphical means for displaying data;
• focus on the schematic face and its computer-based construction; and
• advantages and disadvantages for potential applications of such faces in
the financial environment
Prior to the experiment respondents are issued with sample information sheets
to illustrate the manner in which the statements and ratios will be depicted and
with an illustration of the assignment of ratios to features in the facial
Each respondent is then issued with three separate sets of materials and
asked to make failed/healthy decisions for each of the 20 cases, together with an
indication of the total time spent in processing the materials. Each set of Communication
materials comprises financial information over a five-year period presented in a through cartoon
different medium. For failed companies the fifth year is the last prior to failure. graphics
On completion of the first decision task, respondents are asked to repeat the
exercise successively with 20 sets of company financial ratios and schematic
faces respectively. They are informed that the companies are different in each
Processing orders are varied so that all six possible orderings of statements-
ratios-faces, statements-faces-ratios, etc., are employed. Test materials are
ordered randomly for distribution among the three user groups in order to
determine any impact that order of processing might have on the accuracy and
speed of classification. A common single assignment of financial ratios to facial
features is used throughout. On completion of the experiment respondents
are informed of the identity of the companies and their financial status. Mackay
and Villarreal (1987) express concern over individual differences in cue
responsiveness in the use of schematic faces. They note that content validity
might be lowered because of the comic appearance of the faces, and that females
appear to be more responsive to facial displays than males, both factors having
a potential impact on the quality of the resulting decisions. Although desirable,
the testing of gender effects is not possible with this sample. Only three of the
entire sample of 121 are female, all being MBA students. We might speculate
therefore that the subsequent results might even understate the impact of
Outcomes measuring the “efficiency” and “effectiveness” of the alternative
processing media are detailed in Table I for both the complete sample of 121
respondents and each of the separate user groups. An analysis of these results
highlights two differences, each statistically significant at the 5 per cent level:
(1) The proportion of failed cases misclassified is very much higher than
that of the non-failed cases. This is a potentially important feature of the
Mean percentage of classification errors
Type I Type II Classification time
(Healthy when failed) (Failed when healthy) (Minutes)
Accounts Ratios Faces Accounts Ratios Faces Accounts Ratios Faces
(n = 23) 29.0 31.2 5.1 15.8 29.5 20.8 12.9 11.6 4.0
(n = 52) 34.0 38.8 15.4 12.8 16.6 13.2 11.2 7.6 3.8
(n = 46) 31.2 30.4 9.0 16.0 17.9 16.3 11.4 8.0 4.1 Table I.
Total 32.0 34.2 11.0 14.6 19.5 15.8 11.6 8.5 4.0 Mean error classification
AAAJ results neglected by earlier studies, if, as suspected, the “missed failure”
9,2 is a relatively more important misclassification.
(2) The time spent processing the facial profiles is less than half that spent
on either accounting statements or financial profiles.
Paired-case t-tests are employed to compare the decision-making performance
78 of individual respondents for each of the means of presentation. For each of the
121 respondents the t-statistics generated are shown in Table II together with
the corresponding level of statistical significance.
The faces produce significantly fewer Type I errors, than either the
accounting statements or the financial ratios. The faces produce significantly
fewer Type II errors than the ratios, but not the statements. The high rates of
misclassification with financial ratios are consistent with the findings of
Moriarity (1979) who attribute it to a lack of understanding of what the ratios
Table III shows the corresponding differences for processing times,
demonstrating that the faces are processed significantly more quickly than
either the accounting statements or financial ratios. The facial profiles therefore
Financial ratios Schematic faces
Type I Type II Type I Type II
Type I 0.9 9.1
Type II 4.0 1.1
Type I 10.6
Table II. Type II 2.8
t-statistics for error (0.006)
differences Note: The levels of statistical significance are in parentheses
Accounting statements 6.5 18.1
Financial ratios 13.0
t-statistics for processing (0.000)
time differences Note: The levels of statistical significance are in parentheses
generate decisions at least as good, and often better, than using other sources, Communication
and much more quickly. through cartoon
An analysis of the error classification of Table I demonstrates the extent to graphics
which performance differences are attributable to the accounting sophistication
of the subjects and the processing order of materials. Tables IV and V reveal the
F-statistics, and corresponding levels of significance, resulting from an ANOVA
relating error incidence and processing time to the status of subjects and 79
Table IV shows that order of processing does not significantly influence the
incidence of error for any of the alternative media. The level of accounting
sophistication does impact on classification errors, with the accounting
practitioners making significantly more Type II errors using the financial ratios
Accounting Order of
Type I 0.7 1.1
Type II 2.4 1.8
Type I 2.3 0.9
Type II 6.5 0.4
Type I 5.8 1.4
Type II 5.4 1.2 Table IV.
(0.006) (0.314) F-statistics for Type I/II
Note: The levels of statistical significance are in parentheses errors
Accounting Order of
Accounting statements 1.3 0.4
Financial ratios 7.7 6.2
Schematic faces 0.7 2.7 Table V.
(0.492) (0.023) F-statistics for
Note: The levels of statistical significance are in parentheses processing times
AAAJ and the accounting academics making significantly more Type I and Type II
9,2 errors with the schematic faces.
Observed error patterns help to explain the manner of information
processing and the decision-making strategies employed. A simultaneous
consideration of profitability, short-term debt and balance sheet strength is
sufficient to avoid errors of classification for each of the alternative processing
We may speculate on decision-making strategies employed by reference to
the error patterns generated by respondents. A naïve processing strategy,
applied systematically, of designating companies as “failed” based on a
negative profit before tax figure, generates a familiar error pattern comprising
three Type I errors and two Type II errors. This pattern of Type I errors arises
in 18 per cent of accounting statement misclassifications (and 13 per cent of
financial ratio misclassifications); this pattern of Type II errors arises in 27 per
cent of accounting statement misclassificaitons (and 19 per cent of financial
The frequency of this pattern of errors suggests a myopic profit focus, to the
extent that neither balance sheet information nor industry data receive
appropriate emphasis.This single-variable fixation apparently extends to the
schematic faces, where the most common misclassification pattern (of three
Type II errors, but no Type I errors) is consistent with a strategy of designating
as failed those companies displaying a down-turned mouth. This
misclassification pattern is observed in 26 per cent of error profiles, but the
extent of error is apparently less serious because the schematic faces
automatically incorporate industry averages where a down-turned mouth is
associated not with negative profits, but profitability levels less than the
industry average. Poor performers (potential Type I errors) are, therefore,
identified and the overprediction of failure (Type II error) becomes the dominant
form of error.
Overall, 69 per cent of errors on the accounting statements, 57 per cent of
errors on ratios and 77 per cent of errors on the faces are consistent with a focus
on the profit variable alone. This processing pattern is apparently particularly
prevalent among the accounting academics, resulting in a disproportionate
number of both Type I and Type II errors. The integration of balance sheet
information on the accounting statements and ratios, corresponding to the
incorporation of upper-face features on the schematic faces, allows a rapid
reduction in the number of misclassifications.
Table V shows that the processing time due to the financial ratio analysis is
influenced by both levels of accounting sophistication and processing order.
The accountants take a significantly longer time to process the financial ratios
than either of the other groups, consistent with the unfamiliarity arguments
cited above. The processing time for the ratios is significantly shorter when
they are considered last of the three datasets. The reduction in elapsed
processing time when being processed last is common to all three information
media, but is significantly more marked in the case of the financial ratios.
Discussion and further research Communication
There has been limited study to date of the effectiveness of alternative methods through cartoon
of presenting accounting information for financial decision purposes. This graphics
paper explores empirically the usefulness of the schematic face as a
communication device, in a particular decision context, compared with more
conventional presentation formats. The paper addresses the relative usefulness
of schematic faces, financial ratios and accounting statements as information 81
formats for decision making, demonstrating that schematic faces provide an
alternative means of presentation which might provide a clear and efficient
representation, complementing existing methods. In this respect the paper
provides substantial support for the findings of Moriarity (1979) but in a
manner which produces results far more reliable than those of Moriarity or
Stock and Watson (1984). The Moriarity study has the potential for biased
outcomes favouring faces attributable to accounting ignorance, associated with
the inclusion of too few experimental subjects of any accounting sophistication.
Both of the above studies use arbitrary author-driven selections of feature
assignments which make no reference to the relevant psychological evidence on
the saliency and mobility of facial features. Neither study includes means and
standard deviations for the financial ratio information provided, running the
risk of producing results attributable to superior information and not superior
presentation, given that the schematic faces implicitly incorporate these
statistics. This present study overcomes all of these deficiencies associated with
earlier research in this area.
Evidence is provided that schematic faces are processed more quickly than
either of the more traditional methods of information presentation, with no loss
of accuracy, by users of varying levels of accounting sophistication. Repeated
misclassifications are consistent with overemphasis on the profit figure. Where
the facial profiles produce misclassifications not apparent with the other
processing media, this is consistent with undue emphasis on the mouth as a
facial characteristic. Subject feedback on their response to the use of facial
caricatures, relative to more conventional information forms, is most revealing.
Very little use is apparently made of mean and standard deviation information
when provided, so that the superiority of the face may be at least partly
attributable to the way in which it “forces” subjects to employ this additional
information. The precise assignment of the financial variables to facial features
too, appears to be relatively unimportant once subjects are familiar with the
format. This is reassuring since the message conveyed by the face is so clear
that the opportunities for manipulation might lead us to call for accounting
standards which control the assignment of variables to features. Smith et al.
(1993) suggest that the feature assignment is of much less importance to
processing in practice than the methodology employed. Evidence from
subsequent trials conducted by the authors is consistent with De Sanctis and
Jarvenpaa (1989) who suggest the presence of a learning curve in the use of
graphical information in the accounting profession. Casual observation
suggests that with practice users of schematic faces will develop a holistic
AAAJ perspective and reduce the overemphasis accorded the mouth, with the
9,2 potential for further improved decision making.
The impact of the use of caricatures on the behaviour of subjects can have
both positive and negative effects, with the potential to influence decision
making. The novelty value of schematic faces generates interest and makes them
fun to work with; however, the trend in the use of graphics and other visuals in
82 the annual report, as reported by Beattie and Jones (1992) has not extended to
the use of schematic faces. The message conveyed by the faces may be just too
clear for this purpose, effectively preventing their successful obfuscation by
firms wishing to disguise poor performance. Interestingly in this context,
consulting conducted by the authors to display company schematic faces at the
AGM to convey corporate performance, was discontinued once the faces of
competitors were clearly superior to those of the client company! On the negative
side, the faces may be perceived as trivial and not credible, with the potential for
lowering the content validity of the experiment. One potential subject (among
the group of accounting academics approached here) refused to take part in the
experiment on these grounds. By demonstrating the potential for improved
decision making utilizing new technologies even the most reluctant of
participants cannot help but be impressed; in this case a group of accountants of
a largely conservative demeanour make demonstrably more accurate decisions,
much more quickly, through a medium that they had previously not confronted.
The results presented here demonstrate the usefulness of schematic faces as a
decision tool in the financial environment, with the potential to have a significant
impact on the work of bankers, asset managers and financial analysts. By
providing a speedy, accurate method of processing information, particularly for
extreme cases, the schematic faces may free up management time for the more
detailed analysis of complex situations. These might feasibly include the
performance of investment managers, being appraised simultaneously on a
number of different dimensions of activity, the communication of divisional or
departmental performance based on non-financial achievement, or the
representation of companies by different aspects of their stock market
performance. All of these examples would move away from the failed/non-failed
context, the last away from the good/bad distinction, by searching instead for
patterns of performance that might yield a balanced portfolio.
Future research must also pay more attention to the differences between
individual subjects. Mackay and Villarreal (1987) hypothesize that mental state,
cultural group, personality and psychological factors may be intervening
variables worthy of investigation. Sobol and Klein (1989) echo these concerns,
demonstrating empirically that the efficiency and effectiveness of graphical
displays is dependent on the cognitive style of the respondents. They suggest
that persons with a cognitive style suited to thinking, rather than feeling, have
more success with less traditional graphic forms, though their study did not
extend to a study of schematic faces. These factors should be taken into account
in future studies in this area.
1. The computer code employed to construct Bruckner’s version of the Chernoff face is through cartoon
detailed in Wang (1978, pp. 115-20).
2. The four ratios are:
Profit before interest and tax PBIT
• Profitability by ;
Total assets TA 83
Working capital WC
• Working capital position by ;
Net capital employed NCE
Total liabilities TL
• Financial leverage by ; and
Net worth NW
Quick assets QA
• Liquidity by
Current liabilities CL
These four variables represent the same four dimensions (profitability, working capital,
gearing and liquidity) as employed by Taffler (1982) in his Z-score model. The four
variables above are preferred to his model variables (PBT/CL, CA/TL, CL/TA, NCI) on the
grounds of user familiarity.
3. This visual representation of performance, based on accounting numbers, complements
the idea of “corporate personality” developed in a numerical sense by Sorter et al. (1966).
4. The naïve users comprised a group of 30 first-year undergraduate business students who
had yet to undertake an accounting course.
5. Previous studies (e.g. Houghton, 1984) show that respondents tend to assume an
approximately equal division of failed and healthy companies. Where specific prior
probabilities are indicated (e.g. Libby, 1975) respondents may fail to treat each case on its
individual merits, preferring to rank cases on a best-to-worst basis and then group on the
basis of the given failure base rate.
6. Sample test materials used in the experiment are available from the first author.
7. This is a complex and potentially significant issue. Further empirical work is currently
being undertaken to resolve the specification of an optimum assignment of financial
variables to facial characteristics. Results to date, from Smith et al. (1993), suggest that it is
the adoption of schematic facial profiles which is important to communication, rather than
any particular feature assignment methodology.
8. The total sample of 20 cases is classified correctly through the adoption of an appropriate
linear discriminant model. A simple decision strategy based on a unit-weighted linear
combination of three of the four financial ratios generates one Type II error:
PBIT TL QA
– + < 0.
TA NW CL
A combination of down-turned mouth, small eyes and perplexed eyebrows similarly
correctly identify all of the failed companies from the schematic faces.
Altman, E.I. (1968), “Financial ratios, discriminant analysis and the prediction of corporation
9,2 bankruptcy”, Journal of Finance, Vol. 23, pp. 589-609.
Altman, E.I., Haldeman, R.G. and Narayanan, P. (1977), “Zeta analysis: a new model to identify
bankruptcy risk of corporations”, Journal of Banking and Finance, June, pp. 29-54.
Beattie, V. and Jones, M. (1992), “The use and abuse of graphs in annual reports: a theoretical
framework and empirical study”, Accounting and Business Research, Vol. 22 No. 88, Autumn,
84 pp. 291-304.
Benjamin, M.N. and Pachella, R.G. (1982), “The effects of complexity on interpreting Chernoff
faces”, Human Factors, Vol. 24 No. 1, pp. 11-18.
Bertin, J. (1983), The Semiology of Graphics, University of Wisconsin Press, Madison, WI.
Bruckner, L.A. (1978), “On Chernoff faces”, in Wang, P.C.C. (Ed.), Graphical Representation of
Multivariate Data, Academic Press, New York, NY.
Canadian Institute of Chartered Accountants (CICA) (1993), Using Ratios and Graphics in
Financial Reporting, CICA, Toronto.
Chernoff, H. (1971), The Use of Faces to Represent Points in n-dimensional Space Graphically,
Technical Report No. 71, Department of Statistics, Stanford University.
Chernoff, H. (1978), “Graphic representations as a discipline”, in Wang, P.C.C. (Ed.), Graphical
Representation of Multivariate Data, Academic Press, New York, NY.
De Sanctis, G. (1984), “Computer graphics as decision aids: directions for research”, Decision
Sciences, Vol. 15, pp. 463-87.
De Sanctis, G. and Jarvenpaa, S.L. (1989), “Graphical presentation of accounting data for financial
forecasting: an experimental investigation”, Accounting, Organizations and Society, Vol. 14
No. 5/6, pp. 509-25.
De Soete, G. and De Corte, W. (1985), “On the perceptual salience of features of Chernoff faces for
representing multivariate data”, Applied Psychological Measurement, Vol. 9 No. 3, September,
Diamond, R. and Carey, S. (1986), “Why faces are and are not special: an effect of expertise”,
Journal of Experimental Psychology: General, Vol. 115, pp. 107-17.
Everitt, B.S. (1978), Graphical Techniques for Multivariate Data, Heinemann, London.
Flury, B. and Riedwyl, H. (1981), “Graphical representation of multivariate data by means of
asymmetrical faces”, Journal of the American Statistical Association, Vol. 76 No. 376,
December, pp. 757-65.
Foster, G. (1986), Financial Statement Analysis, Prentice-Hall, Englewood Cliffs, NJ.
Garner, W.R. (1978), “Aspects of a stimulus: features, dimensions, and configurations”, in Rosch,
E. and Lloyd, B.B. (Eds), Cognition and Categorisation, Lawrence Erlbaum, Hillsdale, NJ.
Goldstein, A.G. and Mackenberg, E.G. (1966), “Recognition of human faces from isolated facial
features: a developmental study”, Psychonomic Science, Vol. 6, pp. 149-50.
Grant, E. (1970), “Face to face”, New Society, 7 May, pp. 769-71.
Haig, N.D. (1984), “The effect of feature displacement on face recognition”, Perception, Vol. 13,
Homa, D., Haver, B. and Schwartz, T. (1976), “Perceptibility of schematic face stimuli: evidence for
a perceptual Gestalt”, Memory and Cognition, Vol. 4, pp. 176-85.
Houghton, K.A. (1984), “Accounting data and the prediction of business failure: the setting of
priors and the age of data”, Journal of Accounting Research, Spring, pp. 361-8.
Laughery, K.R., Alexander, J.F. and Lane, A.B. (1971), “Recognition of human faces: effects of
target exposure time, target position, pose position and type of photograph”, Journal of
Applied Psychology, Vol. 51, pp. 477-83.
Libby, R. (1975), “Accounting ratios and the prediction of failure: some behavioural evidence”,
Journal of Accounting Research, Spring, pp. 150-61.
Libby, R. (1981), Accounting and Human Information Processing, Prentice-Hall, Englewood Cliffs,
Lusk, E.J. (1979), “A test of differential performance peaking for a disembedding task”, Journal of Communication
Accounting Research, Vol. 17 No. 1, Spring, pp. 286-94.
Mackay, D.B. and Villarreal, A. (1987), “Performance differences in the use of graphic and tabular
displays of multivariate data”, Decision Sciences, Vol. 18 No. 6, Fall, pp. 535-46. graphics
McKelvie, S.J. (1973), “The meaningfulness and meaning of schematic faces”, Perceptions and
Psychophysics, Vol. 14 No. 2, pp. 343-8.
Moriarity, S. (1979), “Communicating financial information through multidimensional graphics”,
Journal of Accounting Research, Vol. 17, April, pp. 205-24. 85
Morton, J. and Johnson, M. (1989), “Four ways for faces to be special”, in Young, A.W. and Ellis,
H.D. (Eds), Handbook of Research on Face Processing, Elsevier, North-Holland, pp. 49-56.
Reed, S.K. (1972), “Pattern recognition and categorisation”, Cognitive Psychology, Vol. 3,
Remus, S.K. (1984), “An empirical investigation of the impact of graphical and tabular data
presentations on decision making”, Management Science, Vol. 39 No. 5, pp. 533-42.
Rhodes, G., Brennan, S. and Carey, S. (1987), “Identification and ratings of caricatures: implica-
tions for mental representations of faces”, Cognitive Psychology, Vol. 19 No. 4, pp. 473-97.
Ryan, J.A. and Schwartz, C. (1956), “Speed of perception as a function of mode of presentation”,
American Journal of Psychology, Vol. 69, pp. 60-69.
Schaffer, H.R. (1971), The Growth of Sociability, Penguin, London.
Sergent, J. (1984), “An investigation into component and configural processes underlying face
recognition”, British Journal of Psychology, Vol. 75, pp. 221-42.
Smith, E.E. and Nielsen, G.D. (1970), “Representations and retrieval processes in STM recognition
and recall of faces”, Journal of Experimental Psychology, Vol. 85 No. 3, pp. 397-405.
Smith, M. and Taffler, R.J. (1984), “Improving the communication function of published
accounting statements”, Accounting and Business Research, No. 54, Spring, pp. 139-46.
Smith, M., Taffler, R.J. and White, L. (1993), “Cartoon graphics in the communication of
accounting information”, paper presented to the British Accounting Association, University
of Strathclyde, April.
Sobol, M.G. and Klein, G. (1989), “New graphics as computerized displays for human information
processing“, IEEE Transactions on Systems, Man and Cybernetics, Vol. 19 No. 4, July/August,
Sorter, G.H., Becker, S.W., Archibald, T.R. and Beaver, W.H. (1966), “Accounting and financial
measures as indicators of corporate personality – some empirical findings”, in Jaedicke, R.J.,
Ijuri, Y. and Nielsen, O. (Eds), Research in Accounting Measurement, American Accounting
Association, Sarasota, FL.
Stock, D. and Watson, C.J. (1984), “Human judgement accuracy, multidimensional graphics and
human versus models”, Journal of Accounting Research, Vol. 22 No. 1, Spring, pp. 192-206.
Taffler, R.J. (1982), “Forecasting company failure in the UK using discriminant analysis and
financial ratio data”, Journal of the Royal Statistical Society (Series A), Vol. 145 Part 3,
Taffler, R.J. and Sudarsanam, P. (1980), “Auditing the board: a new approach to the measurement
of company performance”, Managerial Finance, Vol. 5 No. 2, pp. 125-47.
Valentine, T. (1986), “Encoding processes in face recognition“, unpublished PhD thesis, University
Wang, P.C.C. (Ed.) (1978), Graphical Representation of Multivariate Data, Academic Press, New
Wilkinson, L. (1982), “An experimental evaluation of multivariate graphical point representa-
tions”, Human Factors in Computer Systems: Proceedings, Gaithersburg, MD, pp. 202-9.
Yin, R.K. (1969), “Looking at upside-down faces”, Journal of Experimental Psychology, Vol. 8,