Table of Contents
1 Introduction - What are Words?
2 State of the Art
3 Data - Priming, Segmenting, Reacting
4 General Discussion - How the Mind Works
5 Conclusion - A Moderate Word-Form Lexicon with Dual-Route Access
1. Introduction - What are Words?
The question of what is stored in the mental lexicon might seem unnecessary at first. “Words, of course.”, one might answer, but: what exactly are “words”? If looked at a little closer, a solution to the problem seems a little more difficult than that. In general, one could differentiate between simple words and complex words. Simple words consisting of only one constituent and complex words having more than one. Sometimes even more than two or three. Furthermore - what are constituents? Linguists all over the world talk about roots and morphemes, differentiating between bound and free morphemes, between prefixes, infixes and affixes and between inflection and derivation. This paper will evaluate how exactly complex words are stored in the mental lexicon and how humans access them. With the help of three major theories, the morpheme lexicon, the strict word-form lexicon and the moderate word-form lexicon, and studies for evidence, this paper finds that one theory, namely the one of the moderate word-form lexicon, might rather apply than any other and that dual-route access is more likely than single-route access. First, this paper estimates how one could understand the mental lexicon in general by then explaining different theories of how complex words are stored in there and how lexical access works. In chapter three, several studies will be listed and explained to be interpreted and combined with the introduced theories in chapter four in order to see which theory goes well with another and which applies more likely. Several examples will be shown to explain certain phenomena, such as regular and irregular past-tense forms. Moreover, this paper will take a look at not just the English language, but also Hebrew, German or Latin to show advantages and disadvantages of the theories.
2. State of the Art
In order to look at what humans store in the mental lexicon it is important to know what this thing is that is called the mental lexicon. Finding a definite brief definition for it can be tough and also it can be hard to describe it, because we are unable to cut open a human's brain to see what it looks like. The mental lexicon is rather an abstract theory and the only option we have is to come up with models to understand the way we process and access words in order to form phrases and sentences for speech. Many models have been introduced, but there are a few things that makes working with them difficult. Those models “are often highly simplified, and [...] often represent guesswork (Aitchison 1994:29). Furthermore, they “can be very different from the original [...]” (Aitchison 1994:29), since they only try to include crucial features that explain certain phenomena. Often, the mental lexicon is referred to as “human word-store” (Aitchison 1994:34), collecting words like a dictionary, which would include the words and much more information about them, such as word class, pronunciation, meaning, orthography and how this certain word is used in a sentence. The best way to understand the mental lexicon is to compare it to things we already understand. The first thing Aitchison comes up with is the London Underground; the mental lexicon as a “diagram of [.] connections [.] we cannot view [.] in the mind directly.” (Aitchison 1994:31). That would mean the words in our mental lexicon are connected and linked like a net of words; some closer and some farther apart from each other and that a person has to follow certain routes (word-links) to form a sentence. But it is questionable if this all is just one giant system or a “set of systems” (Aitchison 1994:31). Kant is also mentioned who compares the memory to a library, in which we label and divide the material like books “on shelves with different labels” (Aitchison 1994:32). In that case, the words in the mental lexicon would liken the books in the metaphor, but here the adjustment of the words is questionable. Ambiguity and frequency could play a role here (Aitchison 1994:33).
But storage in the mental lexicon is not only a question of what we consider the mental lexicon to be, it is also a question of how lexical access works. If we want to know what exactly is stored in the lexicon, we need to know what exactly we activate. Altmann supports the thought that lexicon entries “are not simply accessed [but] they are activated” (Altmann 1997:5) as we search the lexicon for the right word. Haspelmath and Sims (2010) suggest three main theories of how complex words are stored in the mental lexicon (and therefore about what is activated), the first one being that no complex words are stored as such, but that a person only stores base words, so called roots, and morphemes, meaning affixes. This would mean that complex words are basically created by rule by choosing a root and putting it together with the prefix or suffix it needs, for example the word sane is activated and the prefix in-, in order to create the complex word insane. They call this the “morpheme lexicon” (Haspelmath & Sims 2010:63). A second theory is suggested, which contains all complex words to be stored in the lexicon like a “word-form lexicon” (Haspelmath & Sims 2010:63). In that case a person would straight away activate the word insane without any other processes. The third and last possibility is that we indeed store both: all complex words, the roots and the morphemes. They call this “moderate word-form lexicon” (Haspelmath & Sims 2010:63).
What needs to be considered, especially for the last theory, are the differences of single-route and dual-route models. The word-form lexicon would account for a single-route theory, which is basically a matter of full lexical access, meaning a person “follows” only one lexical route to find a word. Eddington (1999) as an example takes regular and irregular inflection. Single-route models support the theory that we access regular and irregular forms with “the same mechanism, either massive storage [...] or equal processing of all forms [...]” (Eddington 1999:282). A dual-route model view would account for different mechanisms while handling regular and irregular inflection, supporting the moderate word-form lexicon. This would mean that a person can “go” both ways: the route from walk to -ed and put it together (decomposition route) and at the same time activate walked (direct route). Which way is quicker “wins” (Haspelmath & Sims 2010:72).
3. Data - Priming, Segmenting, Reacting
It is not always easy to prove a theory right or wrong, but to consider one more likely to apply it is suggested to look at various studies for evidence and see what they found. Prasada and Pinker (1993) looked at the regular and irregular past tense-forms in the English language. The regular past tense rule includes adding the suffix -ed to a regular verb and irregular forms are “unpredictable and hence cannot be generated by a rule” (Prasada & Pinker 1993:2). Using regular inflectional verbs, such as
(1) walk - walked,
irregular not changed past-tense inflection, changed past-tense inflection and past-tense inflection with a vowel change, such as
(2) put - put, sleep - slept, drink - drank (Prasada & Pinker 1993:3)
they performed experiments letting people judge those regular and irregular patterns “by having them rate the [.] goodness and artificial verb stems, and the naturalness of their regular or irregular past-tense form” (Prasada & Pinker 1993:10). Twenty native speakers of English were introduced to sixty novel verbs in six classes and in the end regular past-tense verbs got a higher rating than irregular ones. Even in pseudo-irregular verbs suffixation was preferred (Prasada & Pinker 1993:14). After that they included the verbs in their normal stem-form in a second experiment (Prasada & Pinker 1993:19). The last experiments rather focused on production than rating (Prasada & Pinker 1993:24) and it was still visible that suffixed verbs were preferred, even while considering context in a given blank spaced text.
The study by Feldman, Frost and Pnini (1995) focused on the morphological differences between English and Hebrew and their regular and irregular morphological formations (Feldman, Frost & Pnini 1995:947). In contrast to English, which is a mostly concatenative language using pre- and suffixation, Hebrew is a non-concatenative language. Here, the words have a “skeleton of consonants” (Feldman, Frost & Pnini 1995:948) as their root and more than one morpheme is added via infixation. For the experiments they introduce the segment shifting task, in which “participants are instructed to segment and shift a designated segment from a source word onto a target word [...]” (Feldman, Frost & Pnini 1995:949) and to name the result aloud as fast as possible. In the first experiment, forty American psychology students were introduced to forty English source pairs, one word being complex and the other one simple. During the segment shifting task, for example the word “harden” was given and then the word “bright”. Now the participants had to construct a grammatically correct word with the morphological segments, in this case the word “brighten”. For the participants it was easier to shift segments from morphologically complex words than from words that were non-morphemic (Feldman, Frost & Pnini 1995:951), which suggests that skilled readers are sensitive to morphological structures. The difference in experiment two was that the complex words were now paired with pseudo word-targets, such as the word “harden” paired with the word “eap”, to see whether different word classes play a role. The results were very similar to those in the first experiment (Feldman, Frost & Pnini 1995:952). In the next experiments Hebrew was used to “probe segment shifting in linguistic environments with different morphological characteristics [.]” (Feldman, Frost & Pnini 1995:953). Thirty-six native speakers of Hebrew were introduced to forty stimuli triads consisting of source words, such as devek (glue) and a root like gfn to construct a new word via morphological infixation. The shifting of vowels from transparent roots, that can have more than one meaning was faster than from opaque roots (with fixed meaning) (Feldman, Frost & Pnini 1995:954). In the last experiment they tried the same with source words and pseudo root target words with three consonants. Although the word patterns were more difficult, the results did not differ from those in experiment three (Feldman, Frost & Pnini 1995:956).