The attention demands of lexical retrieval.

Authors Avatar

The attention demands of lexical retrieval

Abstract.

Earlier work has shown that when speakers were asked to name pictures presented simultaneously with conceptually related distracter words their responses on a concurrent tone discrimination task were delayed as much as their picture naming responses (Ferreira & Pashler, 2002).  In contrast, phonologically related distracters had little effect on tone discrimination responses.  From this it was concluded that lemma selection, but not phoneme section is capacity demanding.  By contrast, eye tracking studies show that both lemma selection and phoneme selection are capacity demanding.  The present study aimed to explain the discrepancy between the two studies by firstly replicating Ferreira and Pashler’s experiment.  It was expected that the same pattern of results as in Ferreira and Pashler’s experiment would be found.  However due to faults with the design of the present experiment unexpected results were obtained.  Necessary improvements to the design of the present experiment are discussed.  

Introduction.

 

In producing an utterance speakers must first decide what the utterance should be about, which is considered to be a conceptual and pre-linguistic process.  Secondly, speakers must select appropriate words from their mental lexicon and combine them according to the rules of the language.  The lexical retrieval of a single word involves a number of steps: The retrieval of its meaning and grammatical properties (e.g., whether it is a noun or a verb), the retrieval of its morphological form (where two morphemes would be retrieved for a word such as “walk+ed”) and the retrieval of its phonological form (the consonants and vowels the word consists of).  Most theories of language production assume that only conceptual processes are capacity demanding (i.e. they require the speaker’s attention), whereas lexical access is a fairly automatic process (Levelt, 1989).  The processes involved in lexical access are often assumed to run “in the background,” while the speaker’s attention is focused on planning the utterance contents.  There is some evidence for this view from analyses of pauses, hesitations and speech errors, but there is hardly any relevant experimental evidence.  

When speakers are asked to name two or more objects in noun phrase conjunctions such as ‘the cat and the chair’ (Meyer, Sleiderink, & Levelt, 1998) two types of attention seem to be operating.  One of which is an automatic type of attention, which precedes saccades and is called visual attention.  The other type of attention is more general and is responsible for higher order processes. By registering the speaker's eye movements during the execution of such tasks, eye-tracking studies have found that speakers usually look at the objects they want to name and do so in the order of mentioning them (Meyer et al, 1998). Since a gaze shift corresponds to a shift of attention, this paradigm allows one to discover which of the individual lexical retrieval steps requires attention.

Research using the above paradigm has shown that when naming objects, the speakers eyes remain on the object until the capacity demanding processes are complete, i.e., until they have retrieved the phonological forms of their names (Griffin, 2001; Meyer & van der Meulen, 2000; Meyer et al, 1998).  The higher order attention is then able to shift to a new object.  Since people usually look at the objects they attend to, this suggests that lexical retrieval processes are more capacity demanding than commonly assumed.  Alternatively, the capacity demanding processes may not be the lexical access processes themselves but accompanying speech monitoring processes.  It is apparent from spontaneous self-repairs that speakers can monitor their ‘internal speech’ (Levelt, 1983).  There is strong experimental evidence that the representation that is accessed in speech monitoring is the phonological representation of the utterance (e.g., Wheeldon & Levelt, 1995).  Perhaps these speech monitoring processes require conscious attention.

Dual-task reaction time paradigms, which are widely used by attention researchers, have recently been employed to study which processes involved in speech planning require central processing capacity.  Attention studies have demonstrated that when participants are required to perform two reaction time tasks concurrently, responses to the second stimulus are delayed (see Meyer, D & Kieras, 1997; Pashler, 1994, 1998, for reviews).  This form of dual-task interference, known as the psychological refractory period (PRP) effect has been accounted for in a number of ways.  One possibility is that when two tasks require central processes at the same time, a bottleneck results.  Assuming that Task 1 claims the bottleneck mechanism first, central processes needed in Task 2 must be postponed until the corresponding processes of Task 1 have been completed (Pashler, 1994).  

To test which of the individual lexical retrieval steps are subject to the central processing bottleneck, Ferreira and Pashler (2002) asked participants to perform two unrelated reaction time tasks concurrently.  For Task 1, participants were required to name pictures and for Task 2 they were required to discriminate between three tones.  Earlier studies using similar paradigms have shown that the tone discrimination task requires central resources. The central processing bottleneck model predicts that if Task 1 also requires central resources, the tone discrimination latencies will depend on the difficulty of that task. In Experiment 1, Ferreira and Pashler manipulated the difficulty of lemma and phonological word-form selection with high and low frequency name pictures and high and low constraint sentences. Earlier studies have found that word frequency affects the retrieval of a words phonological form (e.g., Jescheniak & Levelt, 1994) whereas contextual constraint effects lexical selection (Griffin and Bock, 1998).  

Join now!

Ferreira and Pashler found that pictures with high frequency names and pictures preceded by high constraining sentences were named faster than pictures with low frequency names and pictures preceded by weakly constraining sentences. They also found that the tone discrimination latencies were affected by these semantic and morphological manipulations.  The longer the semantic and morphological processes took, the later participants responded to the tone.  This suggests that concurrent processing cannot occur during lemma selection or during morpheme retrieval, a finding that converges with observations of eye movements during lexical access reported by Meyer et al (1998) and Meyer & ...

This is a preview of the whole essay