Behavioral studies of human language
Today, there is a wealth of information about human language and the underlying neural mechanisms. A lot of that information has been accumulated in studies on language deficits exhibited by neurological patients with focal brain lesions and, more recently, in non-invasive neuroimaging studies with healthy volunteers. Prior to development of modern non-invasive neuroimaging methods, however, behavioral experiments in healthy volunteers have been carried out in a rather ingenious manner to reveal important aspects of human language functions. Examples of how behavioral methods have been utilized in human language studies are described in the following.
Mental lexicon
The vocabulary of the human language (also called “mental lexicon”) is highly excessive with, depending on the language, the amount of words ranging from tens of thousands to hundreds of thousands. The relative size of the vocabulary depends on how one counts and the nature of the language in question, for instance, whether only root words are counted (in which case so-called isolating languages appear to be larger than other languages) or whether one counts only words that are currently used. Nonetheless, vocabularies of each of the human languages are extremely rich, yet one is able to access
each word relatively effortlessly, given that the words are produced and comprehended with an approximate rate of three per second (there is, of course, variability between languages in the average length of a single word).
The organization of the mental lexicon has been a topic of numerous behavioral studies (as well as that of neuroimaging studies, as will be described later in this chapter). As a simple example of how recording of reaction times could be used to study mental lexicon is measurement of reaction times to words that begin with different letters of the alphabet to answer the question of whether the mental lexicon is organized in an alphabetical order like any ordinary dictionary. If this were true, one would expect longer reaction times to words as a function of serial position within the alphabetically organized dictionary (i.e., longer reaction times for words that begin with letter “w” than for words beginning with letter “a”). In reality, of course, the mental lexicon is not organized in an alphabetical order.
Rather than alphabetic organization, it has been observed that words which are more frequently used in daily life are accessed faster than less frequently used words (Forster and Chambers, 1973). Incidentally, it also seems that the more frequently used words are the ones that evolve slowly. For instance, across European languages the words for “two” are rather similar. In contrast, the words for “tail” are rather different (Pagel et al., 2007). In addition to frequent words, words that are more unique in how they sound are accessed faster than words that are acoustically less unique (e.g., words “eight”, “late”, and “rate” sound alike; homophones such as “too” and “two” are the prime example of this).
Semantic networks of words
Behavioral studies have suggested that words which are semantically close to one another also seem to be close to one another at the level of brain representations. These so-called priming studies are based on a relatively simple experimental design where two words are presented in close succession. The first word is a prime and the second word is a test word. The test word is either a real word or a pseudoword (i.e., a string of letters that resembles a real word but does not mean anything such as “tacke”). The task of subjects is then to press, as quickly as possible, a reaction time button if they judge the test word to be a real word. If the first word is semantically close to the test word, the reaction times are significantly faster than if the two are semantically unrelated (Neely, 1977). For instance, truck-car pairing results in faster responses than rose-car pairing.
Based on results from these studies, it has been possible to construct semantic networks of words (see Figure 9-1), where semantic concepts are nodes of the network and there are links between the nodes representing semantic associations between the nodes. The distance in the network (i.e., number of links between two concepts) reflects distance in the semantic space. It is assumed that presentation of a word or a concept results in spreading of activation across the semantic network, thus explaining the priming effects. For an early review on the behavioral studies on semantic language networks, see (Collins and Loftus, 1975).