Sie sind auf Seite 1von 12

Psycholinguistic Principles in the Game Taboo

by Jeffrey Thomas Diteman


Portland State University, 2013

Taboo by Hasbro is a word guessing game in which a few simple constraints are
imposed to make the task more challenging. Unlike some other guessing games, it is strictly
verbal: no drawings, gestures or spellings are involved. Therefore, the rules and behaviors
found in Taboo are related to psycholinguistic principles that pertain mainly to semantics.
Knowledge of semantic theory is useful to understand the game, and an examination of the
game provides insights about how people use words and the relationships between them to
convey meaning.
The rules of Taboo are straightforward. Players divide into two teams. A clue-giver
tries to get the rest of the team to guess the target word. The target word is found on a card
which also contains a list of forbidden words, which the clue-giver must avoid using in
attempting to describe the target word. During a turn, the team tries to guess as many
words as they can, based solely on the clue-givers description, within a limit of one minute.
The clue-giver can pass on cards that seem too difficult. When the minute is up, play passes
to the other team; teams gain one point for every card correctly guessed and cede one point
to the opposing team for every card passed, and for any other error. Errors include saying
any part or form of a word found on the card, making gestures or sound effects, and giving
hints about the sound or spelling of the target word.
The game relies on the psycholinguistic process of semantic access. The fun of the
game comes from attempting to convey meaning without using the usual tools. In playing,
the normal processes of semantic communication are modified, as the clue-giver is
prohibited from using the most obvious types of semantic and lexical information about the

target word. The forbidden words are those most closely associated with the target word,
either by meaning or by co-occurrence. Since no clues about the forms of the words are
allowed, the route to access in Taboo is strictly semantic. Considering that no gestures,
pictures or other reference clues are allowed, the players must rely on the sense of words;
only in isolated cases will shared private stories facilitate any reference-based
communication.
In order to understand how the game works, it will be useful to understand the
process of semantic access. Accounts of semantic access seek to explain the organization of
the lexicon and the phenomena observed in human speech. Theorists look for evidence for
the psychological reality of explanations for the surface forms and strategies found in daily
communication. For example, can it be said that there is a psychologically real hierarchical
organization of the lexicon? Under the lens of reason, words can be organized hierarchically.
Most open-class, lexical words (as opposed to closed-class, grammatical words) are related
to others such that any basic-level concept may have various subordinate and superordinate
concepts, along multiple axes of conceptualization. For example, if car is taken as a basiclevel word, a superordinate category is motor vehicle and subordinates include coupe,
sedan, station wagon, etc. Of course, this is only the subordinate set corresponding to body
types; other subordinate sets include makes and models, sets defined by number of
cylinders, type of transmission, and so forth. Furthermore, motor vehicle is not the only
conceivable superordinate; all superordinates of the set motor vehicles are also
superordinates of car. Therefore, a car is a vehicle, and a thing with wheels, and a
manufactured item, etc.
Another way to think about semantic organization is in terms of types and tokens. If
sedan is taken as a type, the Ford Taurus, Toyota Camry and Chevrolet Cavalier are tokens of
that type; they are specific subordinates of the category sedan. Again, this is not the only
possible level for tokens of the type. In looking at a parking lot, we could pick out individual

instances of sedans, and in so doing, we might find several individual Ford Tauruses; for the
purposes of the task, they are merely sedans like any other: their similarity to each other
has no bearing on their class membership if all sedans are considered equal. Conversely,
another task could call for ranking individual cars in terms of their goodness of fit to the
category sedan. For the purposes of the exercise, some will be more prototypical than
others, so the Tauruses might be treated differently from the Cavaliers, even though they
both meet the basic criteria.
All of these possible configurations contribute to the nuances of meaning we find in
everyday speech. The lexicon can be said to be organized hierarchically, but in a flexible
way: different contexts call for different organizational strategies. Players of Taboo can
employ this flexibility of language to work around the constraints imposed by the rules of the
game.
Any theory of semantic access must account for the phenomena observed in human
speech. Some phenomena are readily observable while others are detected through
controlled experiments. We know that semantic access must be very fast, because average
speakers are able to utter and comprehend two to three words per second in fluent speech
(Santelmann, 2012a). Furthermore, word processing is automatic. This means that upon
hearing a given word, a person associates it with the concept it represents without conscious
effort, and nearly instantaneously. In addition, observation suggests, and research
demonstrates, that perception of any given word produces automatic activation of other,
related words. The activation of related words can be tested in experiments that evaluate
word priming effects using naming or lexical decision tasks. Exposure to one word causes
faster response times for the production or recognition of related words. Thus, priming a
subject with the statement a canary can fly will lead to a faster response time when the
subject is subsequently asked to assess the validity of the statement a canary is a bird

(Traxler, 2012). This suggests that the notion of bird was already partially activated in the
subjects mind.
One of the greatest challenges in semantics is to explain how word meaning
transcends dictionary-style definitions. A given word can have a wide range of nuanced or
even contradictory meanings depending on its context. The old debate about whether a
tomato is a fruit or a vegetable is only the beginning. In the kitchen, a tomato is treated as a
vegetable; to a botanist, it is the fruit of the tomato plant. Notice the semantic quagmire
already developing: the word can also be used to refer to the whole plant, or to the species.
In given situations, the word could be used to refer to the seeds of the plant, or to an entire
patch of plants, or, for that matter, to the flavor of tomatoes, a shade of red, a sunburned
person or a voluptuous woman.
In light of the difficulty of describing words in terms of their definitions, the history of
semantic theory has been characterized by a shift from feature-based to network models.
Attempts to describe word meaning in terms of atomic features have a limited ability to
explain the semantic possibilities and phenomena. As outlined in Santelmann, 2012b, initial
feature models proposed a decompositional approach, whereby words are broken down into
basic units of meaning. One way in which the feature-based models were refined to account
for the lability of meaning was in distinguishing between defining and characteristic
features. Another improvement involved ranking features for salience and probability. For
example, one might say a tomato is a vegetable because its most salient and probable
features, its flavor and its use in savory dishes, place it in that category. Finally,
incorporating the notion of prototypicality can explain why some items class pertinence is
easier to define than others. For example, a tomato can be a fruit (it is the seed-containing
berry that develops from a fertilized flower), even though it is a non-prototypical one (it is
not as sweet as most edible fruits). Nevertheless, the fact that nearly everyone considers it
a vegetable and not a fruit is an example of one of the major problems with feature theory:

how can something that meets the defining criteria for fruithood not be a fruit? Conversely,
we can think of examples of fruits that do not meet the criteria for fruithood: growers have
selectively bred seedless grapes and watermelons into existence. Containing seeds, which
appeared to be a common and necessary criterion for class membership, is no longer; we
would be hard pressed to find someone who would claim that seedless grapes are not, in
fact, fruits. All fruits have seeds, except for those that do not, just as all mammals give live
birth, except for those that do not, and so forth.
The vagueness and subjectivity of feature-based definitions have led many theorists
to opt for a semantic network model instead of feature-based models. According to semantic
network theory, a words meaning comes from its position as a node in an array of nodes
connected by semantic links. The shift is away from the concepts features and toward its
associations; synonymy, antonymy, hierarchical pertinence, utility, co-occurrence, and many
other types of association are possible. This theory accounts for superordinate-subordinate
hierarchical structures through the principle of transitive inheritance: a tomato is a
vegetable, and vegetables are plants, therefore a tomato is a plant. Powerfully, if the lexicon
is modeled as an interconnected web of semantic links, class membership can be
reconfigured extemporaneously, depending on context. Furthermore, some words have
higher connectivity than others, explaining why some words are easier to access.
In terms of processing, semantic network theory is based on a principle of spreading
activation (Traxler, 2012). This principle explains the speed and automaticity of access and
priming. Hearing, reading, or thinking of a given word produces partial activation of related
concepts. This would explain the experimentally demonstrable priming effects. Furthermore,
activation is conceived of as dissipating as it spreads, so only concepts within a limited
number of degrees of semantic relatedness are activated. Nevertheless, the activation of
multiple degrees makes possible the phenomenon of mediated priming, whereby, e.g., the
word lion primes the word stripes via the intermediary word tiger.

Lexical hierarchies, semantic networks and spreading activation are useful to explain
the processes at work when people play Taboo. The game makes use of our ability to
convey concepts indirectly. The clue-giver is required to guide team members to the target
word in a context where the usual pathways of meaning and form are not allowed. No
gestures or sound effects are allowed, so the clue-giver cannot rely on direct embodied
information. There are rules against giving any clues about the form of words: no rhymes
and no initials are allowed. No variant or part of any of the target or taboo words can be
used. The taboo words on any given card are some of the most common associates of the
target word, related to it by meaning, frequent co-occurrence, or both. In fact, the target and
its taboos can be arranged in a classic semantic network diagram:

grape

is

wrinkled
appea
rs

is

RAISIN

resembles

prune

dried

grown in

California

The forbidden words may be subordinates, superordinates, co-occurring terms,


conceptual correlates, or features of the target word. That features are only one element
among many is evidence that meaning transcends features. For raisin, +grape, +dried and
+wrinkled are all features; grape can be thought of as a superordinate, since all raisins are
grapes, but not all grapes are raisins. [+Grown in California] could be construed as a
probable but not defining feature of raisin, but California is better described as a cooccurring and conceptual associate. Prune is an associate related by conceptual similarity

but not necessarily co-occurrence. Its inclusion on the card, and in this semantic network,
illustrates the pertinence of semantic associates that are neither features nor co-occurring
lexical items. If prune were not forbidden, talking about a fruit that resembles a prune would
be a quick and effective way to convey the notion of raisin. This is evidence of the
psychological reality of semantic networks, interconnected by links that can have a wide
variety of conceptual values. Semantic relationships of similarity like this one can be
evaluated using picture naming tests. Presenting subjects with pictures superposed with
mismatched words from the same semantic category causes more interference with naming
them than if the non-relevant word comes from a different semantic category. That is,
people have a harder time naming a picture of a dog when it is labeled cat than when it is
labeled juice (Traxler, 2012).
Some insights from feature theory can be useful in understanding the taboos. For
one, the features that are forbidden are the most salient and probable features: raisins are
quite obviously, and almost always, dried and wrinkled. That they have a higher sugar
concentration and therefore taste sweeter than grapes is a quasi-universal feature, but,
being less salient, does not apply. In fact, a player attempting to use such less-salient
features to get around the taboos is likely to find limited success, depending on the
preconceived notions and mental habits of his or her teammates. A feature-based strategy
for this target might go, This is a small, sweet, purple fruit. It is chewy. It is yummy in
oatmeal. Such strategies are not particularly effective, and are in fact infrequently used by
actual players. A clue-giver is much more likely to employ strategies based on synonyms,
conceptual relatedness, co-occurrence and shared knowledge. For example, for the target
word raisin, one player in actual game play said This is a food, in the nineties there were
these commercials with four dancing, singing guys, shaped like and this was sufficient for
a teammate to shout the answer. The way this player makes an appeal to a shared pop

culture reference shows how association, and not merely pure meaning, plays an important
role in semantics.
Given the total block on form-related clues and the obstacle prohibiting the most
immediate lexical neighbors, direct paths to the word are blocked. Therefore, players must
rely on indirect strategies. These can take the form of synonyms and associations; degrees
of separation that evoke the taboo words in order to establish the domain; or personal
anecdotes or other forms of shared knowledge.
In adult players, the use of superordinate synonyms is a common work-around. Take,
for example, the target word bayonet, for which the taboo words are rifle, sword, stab and
soldier. A player had quick success by saying The sharp thing at the end of a gun, using
the superordinates sharp thing for knife and gun for rifle. This is one of the simplest
strategies. Since it will not work for all targets, players must be flexible in their approach to
communication. The target card family tree forbids the word ancestors but does not forbid
the word descendants, so a fast-thinking player will be able to use the quick work-around of
the antonym in order to convey the idea.
In order to get the team thinking of the target domain, some players use
descriptions, rather than synonyms, for the taboo words. To get around the taboo word Italy
in attempting the target word spaghetti, a clue-giver said, This is a food from that comes
from that boot-shaped peninsula on the south side of Europe. This corresponds to
conceptual activation of Italy using a general description of its salient features. It would have
been possible to accomplish a similar feat by going more specific: This is a common food in
Rome and Naples. The fact that Rome and Naples automatically prime Italy and indirectly
prime spaghetti is evidence of spreading activation at work. It would be interesting to
examine which strategy for establishing the domain is faster and more consistently
successful in actual play.

Phenomena observed in game play can tell us about how semantic access works in
time. For example, the following exchange was observed for the target word orthodontist:

Clue-Giver: This is the guy who makes sure that the things that you chew with are straight.
Guesser: Dentist. Orthodontist.

The guesser was first led to dentist by the first part of the clue, and only after processing the
mismatch between that concept and are straight stumbled on the correct answer. This is
evidence that speech is processed incrementally.
Furthermore, when it comes to compound words, conceptual descriptions that lead
directly to their subunits are sometimes more successful than those which seek to elicit their
target concept. When cluing for the target word homepage, one player used the following
strategy:

Clue-Giver: When you wake up and turn on the thing that wakes you up that has a screen
Guesser 1: Phone!
Guesser 2: Computer!
Guesser 1: Internet!
Clue-Giver: We live at
Guesser 1: Home! Homepage!

This is an example of a two-stage cluing process, which first defines the domain and then
narrows in on the target by cluing for a subunit of the target. The fact that the guessers tend
to start calling out multiple words from the domain is evidence of automatic priming and
spreading activation at work. The clue-givers dynamic response to the other players, taking
into account that they had the right domain but needed more specific guidance, illustrates

the flexibility of cognition for communication. Semantic network theory accounts for the
relative ease of access of more common words by asserting that higher frequency of use
leads to greater connectivity; high-connectivity words like home are easier to access than
low-connectivity words like homepage (Traxler, 2012). Since the notion of home is much
easier to convey than homepage, the player chose it as a way to narrow in from the domain
of computers to the specific target. Such cross-indexing strategies are common in game
play, and tend to be met with success.
Another effective strategy is to employ tokens to represent types. The drugstore card
prohibits the synonym pharmacy, but does not prohibit the token Rite-Aid. The use of tokens
for types relies on shared knowledge that goes beyond simple semantics, requiring
knowledge of the world. A very young player might not think of Rite-Aid as a drugstore; to a
4-year-old, it seems more like a candy store. Therefore, a shared understanding of the class
of the token is necessary. The use of tokens for types also applies to the strategy employed
by the player who invoked the clay-animation characters, the California Raisins, in order to
convey the notion of raisin. A specific cultural reference can bring the other players
immediately to the target word through this sort of association.
This brings us to one of the main critiques of the semantic network theory: the
symbol-grounding problem. If words are considered only in terms of their relationships to
other words, they may as well correspond to an array of meaningless symbols. Only through
their association with our knowledge of the way things work in the world to words have true
meaning. For this reason, embodied semantics asserts the indexical hypothesis, which states
that the first step in meaning is making the connection between a word and its real-world
referent or to analog mental representations of the referent. When it comes to
communication, shared knowledge, whether cultural or personal, makes it possible for
people to have some confidence that their respective lexicons map, at least roughly, to the
same concepts.

There are exceptional cases in Taboo when specific references come into play.
These are those rare instances when a word happens to overlap with the players recent
shared experience. For example, if the word lasagna comes up, and two of the players had
shared a lasagna dinner that week, there will be an opportunity to exploit this anecdote in
order to arrive at the answer quickly. This is another example of a grounded symbol, and
requires a theory of semantics that is expanded to incorporate the indexical hypothesis.
Such a referential phenomenon can also arise within game play itself, such as if the card
ballet is drawn shortly after the word step. The clue-giver will be able to exploit the fact that
other players have already been primed for ballet by referring back to the previous card.
Such cases are different from the sort of cultural reference found with the California Raisins,
in that they employ specific instances in the players shared experience. These referential
cases bring the play of Taboo beyond the scope of semantics and into that of discourse.
This is a further illustration of the flexibility of meaning, which makes it difficult for any one
theory to account for all aspects of semantics.
In conclusion, Taboo is a game in which cleverly devised constraints exploit and
modify the usual processes of semantic access to convey meaning. The linguistic principles
on which the game is built make it primarily a game of semantics. It relies on players shared
semantic knowledge, and to some extent on their shared cultural or personal knowledge. In
addition to its amusement value, Taboo could be useful in assessing and developing
communication strategies in young players and second-language learners. In order to make
the game more appropriate for the L2 population, cards such as Dr Laura Schlessinger that
rely on highly specific cultural knowledge could be omitted so that the more strictly
semantic aspects of play can come forth. The ability talk about a concept indirectly is one of
the skills that L2 teachers try to cultivate in their students, so that conversations can be
maintained in the target language without falling back on translations from L1. Therefore, a
modified version of Taboo might make an effective activity in the advanced L2 classroom.

Such activity-based learning can reinforce the connectivity of words in the semantic
network, promoting quicker access while expanding vocabulary.
References

Santelmann, L. (2012). Speech Perception: Sounds [Powerpoint Slides]. Retrieved from:


https://d2l.pdx.edu/d2l/lms/content/viewer/main_frame.d2l?ou=325442&tId=1276730

Santelmann, L. (2012). Word Processing [Powerpoint Slides]. Retrieved from:


https://d2l.pdx.edu/d2l/lms/content/viewer/main_frame.d2l?ou=325442&tId=1276731

Taboo Instructions (2000). Pawtucket, RI: Hasbro.

Traxler, M. (2012). Introduction to Psycholinguistics: Understanding Language Science. Malden, MA:


Wiley-Blackwell.

Das könnte Ihnen auch gefallen