Isolating Phonological Components 1 Running Head: PHONOLOGICAL PRIMING AND TIP-OF-THE-TONGUE Memory and Cognition, 31, 1153-1162 Isolating Phonological Components that Increase Tip-of-the-Tongue Resolution

Three experiments investigated the role of specific phonological components in priming tip-ofthe-tongue (TOT) resolution. When in a TOT state, participants read a list of words that included phonological primes intermixed among unrelated words. The phonological primes contained either the same first letter as the target (Experiment 1), a single syllable (first, middle, or last) of the target (Experiment 2), or the first phoneme or first syllable of the target (Experiment 3). Reading first-letter primes in Experiment 1 did not help to resolve TOTs, whereas reading firstsyllable primes significantly increased word retrieval in Experiment 2. Experiment 3 replicated the results of Experiments 1 and 2 using first-phoneme primes instead of first-letter primes and using two primes instead of three. Experiment 3 also extended first-syllable priming to primes read silently. The results of these experiments support a Transmission Deficit model, where TOTs are caused by weak connections among phonological representations and can be resolved through internal or overt production of specific phonology. Abrams Isolating Phonological Components 3 Isolating Phonological Components that Increase Tip-of-the-Tongue Resolution A tip-of-the-tongue state (TOT) occurs when a word is temporarily inaccessible for retrieval, despite a feeling of knowing the word. Although the cause of TOTs has been debated (see A. Brown, 1991, for a review), one predominant view is that the target word does not receive sufficient activation to allow for word retrieval (e.g., R. Brown & McNeill, 1966; Meyer & Bock, 1992; Yaniv & Meyer, 1987). A more specific account comes from the Transmission Deficit (TD) model of TOTs (Burke, MacKay, Worthley, & Wade, 1991; James & Burke, 2000; MacKay & Burke, 1990; Rastle & Burke, 1996), which states that TOTs are caused by weak connections between lexical nodes and phonological nodes that are represented in a hierarchical network of multi-level systems (see Burke et al., 1991; MacKay, 1987; MacKay & Burke, 1990, for a description of this network called Node Structure theory). The present experiments tested this account of TOTs by strengthening connections to specific phonological nodes and assessing the effects on TOT resolution. These experiments investigated which phonological features were capable of facilitating word retrieval by priming specific phonological features, i.e., the first letter of the target word, the first phoneme, or a single syllable, during TOT states. Node Structure theory and the TD model offer a specific explanation for why TOTs occur and how they may be resolved by postulating two processes that determine word retrieval: node priming and node activation. Node priming is a form of excitation that is transmitted across connections between nodes, preparing them for activation (other network models refer to this as “spreading activation”, e.g., Anderson, 1983; Collins & Loftus, 1975). Node activation is an “all or none” process that occurs when the level of node priming exceeds some threshold, which is required for word retrieval. According to the TD model, a TOT state occurs when the connections to a word’s phonological nodes are weakened, causing a reduction in the Abrams Isolating Phonological Components 4 transmission of node priming to these nodes. Without a sufficient amount of node priming, these phonological nodes cannot get activated, and therefore word retrieval cannot occur. Connections to a word’s phonological nodes weaken over time if they are not activated frequently or recently. For example, if the lexical node for the word contraband has not been activated often in several years, a TOT is more likely to occur when trying to retrieve this word because the connections to its phonological nodes, e.g., the sounds /kon/, /tr /, and /band/, are weakened. However, some aspects of phonology, such as the first letter or phoneme, can be retrieved during a TOT state in spite of the inability to retrieve the entire word (Burke et al., 1991; Harley & Bown, 1998; Meyer & Bock, 1992; Miozzo & Caramazza, 1997). Within the TD model, this “partial retrieval” occurs because sufficient node priming accumulates on a particular phonological node and leads to activation of the corresponding phonological component. For example, the first phoneme of the word contraband may get activated and result in the feeling that the word begins with “c”. However, for complete word retrieval to occur, every phonological node must be activated. Previous researchers have attempted to activate the missing phonology and increase TOT resolution by implementing an intentional cueing paradigm (e.g., Bowles & Poon, 1985; Jones, 1989; Jones & Langford, 1987; Maylor, 1990; Meyer & Bock, 1992; Perfect & Hanley, 1992). For example, Meyer and Bock (1992) presented participants having a TOT with a word that shared multiple phonological features with the TOT target word (e.g., the initial sound and letter, number of syllables, and stress pattern) and informed them that the cue “may help you to think of the word” (p. 719). They found that more TOTs were resolved when participants were presented with phonological cues than semantic or unrelated cues, suggesting that phonological cues facilitate retrieval when in a TOT state, as predicted by the TD model (see Jones, 1989, and Abrams Isolating Phonological Components 5 Jones & Langford, 1987, for exceptions). However, facilitation from cues in their study most likely reflected strategic retrieval processes. Knowing that the cue may be related to the target word can encourage a directed memory search for the target, using phonological properties of the cue. Although cueing produces facilitation in word retrieval, the nature of this facilitation is different from the facilitation produced by automatic retrieval processes that occur during a TOT state. To minimize the influence of conscious retrieval strategies on word retrieval when in a TOT state, James and Burke (2000, Experiment 2) employed a type of modified repetition priming paradigm that disguised the relation between the primes and the TOT target word. After indicating that they were in a TOT state, participants pronounced aloud and rated the pronunciation difficulty of ten words, five of which cumulatively contained all of the target word’s phonology. The results showed that participants resolved more TOTs following phonological primes than unrelated words. Even though these results are consistent with cueing paradigms, they are more likely to accurately reflect the automatic processes involved in normal word retrieval (i.e., words “popping” into mind instead of using conscious search strategies). In terms of Node Structure theory and the TD model, James and Burke (2000) concluded that a relatively long-term change in connection strength (which occurs without an individual’s awareness) is responsible for increased target word retrieval in the phonological priming paradigm. Similar to James and Burke (2000), the present experiments assessed TOT resolution following presentation of phonological primes, not cues. However, our experiments primed participants with words that contained specific aspects of phonology (e.g., first letter/phoneme or a single syllable), unlike James and Burke (2000) and cueing studies that have used words that Abrams Isolating Phonological Components 6 shared several phonological characteristics with the target word. The present experiments are the first to explore the effect of isolated phonological features on TOT resolution, independent of conscious retrieval strategies. Node Structure theory and the TD model can be used to predict the aspects of phonology that are most likely to facilitate TOT resolution. Specifically, they predict that the critical phonological aspects necessary for word retrieval are likely to be located in the initial portion of the word. In Node Structure theory, language production involves sequential left-to-right activation (MacKay, 1987): One must activate the phonological nodes at the beginning of a word before activating the nodes in the next segment, and so forth until the entire word is activated and can be produced. If the initial segment fails to get activated, then subsequent portions of a word cannot get activated, which then results in a TOT. Furthermore, the sequential nature of activation suggests that the initial portion of the word is also critical for cueing or priming word retrieval, such that strengthening connections to a word’s initial phonological nodes is necessary for activation of the remaining segments when in a TOT state. This prediction has been supported by research on cueing with different parts of target words, which has shown that the initial part of a word, e.g., first letter or sound, serves as the best retrieval cue (e.g., Bowles & Poon, 1985; Freedman & Landauer, 1966; Gruneberg & Monks, 1974; Heine, Ober, & Shenaut, 1999), especially in comparison to middle or end letters (Horowitz, White, & Atwood, 1968). In contrast to activation of the initial segment, failed activation of a word’s middle or final segments is less likely to have an effect, either in causing TOTs or in facilitating TOT resolution. Even if the latter segments do not get activated immediately, the initial phonology that is available can be used to retrieve the remaining phonology and enable word retrieval, preventing a TOT from occurring. In terms of TOT resolution, activation of the middle or final Abrams Isolating Phonological Components 7 segments during a TOT does not ensure that the initial segment will get activated, resulting in no increase in word retrieval following cues or primes containing the latter syllables. Latter syllables are also likely to be less effective primes than initial syllables because of inherent characteristics of syllables at the end of words. Certain parts of speech often have similar endings (e.g., many nouns end in the suffix “tion”, and many verbs end in the suffix “ate”), which make them more predictable. Previous research supports this claim by demonstrating a greater-thanchance probability of guessing the final letters of words when in a TOT state (e.g., Kohn et al., 1987; Koriat & Lieblich, 1975; Rubin, 1975). Because many words share these endings, the latter syllables will be used more often, and their higher frequency of use will make the connections to their phonological nodes stronger and less susceptible to deficits in the transmission of node priming across these connections. However, unless the firstand middlesyllables are activated, the last syllable will not be retrieved. Consistent with these predictions and the partial information that people often report when in a TOT state, we designed our studies to test the effects of first letter (Experiment 1), a specific syllable (Experiment 2), and first phoneme versus first syllable (Experiment 3) on TOT resolution. Experiment 1 In this experiment, we investigated whether primes with the same first letter as the target (and same first phoneme 82% of the time) can facilitate TOT resolution, when the use of conscious strategies to facilitate retrieval are minimized. As discussed earlier, Node Structure theory postulates that activation is a sequential process; therefore, the initial sound of a word must be activated before later sounds can be activated. Consequently, providing the first letter should activate the initial sound, boost the transmission of priming to the other phonological nodes, and facilitate resolution of the TOT. Abrams Isolating Phonological Components 8