1. Compound elicitation in Finnish: the case of writing Bertram, R. Toennessen, F., Strömqvist, S. Hyönä, J. and Niemi, P. Poster presentation during the 45th annual meeting of the Psychonomic Society, 18.11-21.11, Minneapolis, USA, 20.11, 12.00-13.30

2. In psycholinguistics, from all four modalities, writing is the least studied and therefore least understood one. This does not imply that scholars think of this modality as the least interesting one, but rather as the modality that is the hardest to study. Indeed, there are very few tools that allow one to get an insight in the writing process in general and even less to get an insight in the on-line writing process. Recently, Strömqvist and Malmsten (1998) developed a tool, coined Scriptlog, which does exactly that. In other words, Scriptlog allows one to investigate all kind of important aspects of writing behavior. The program registers practically everything: the production time from letters to whole texts, typing errors and their corrections, pauses, transition times between different linguistic units, etc. Consequently, it can be used to investigate writing behavior on all linguistic levels: the word, sentence and text level. Writing behavior can be elicited by pictures or text or by simple oral instructions. Background

3. The present compound study In the current study, we investigated the role of morphology in writing words. In one experiments, including two experimental manipulations, participants had to write down pictures names by means of a keyboard. The target pictures were supposedly eliciting compound words that either varied with respect to first constituent frequency or whole word frequency. lento/kone ’airplane’ high-frequency whole word tuhka/kuppi ’ashtray’ low-frequency whole word tasku/lamppu ’flashlight’ high-frequency 1st constituent silitys/rauta ’ironing iron’ low-frequency 1st constituent

4. The present compound study Frequency is often used as a diagnostic tool in studying complex words: the frequency of the constituent morphemes or that of the whole word is manipulated in order to assess whether morphological substrings and/or whole-word forms are employed in the course of processing. We assume that if low first-constituent frequency compounds have longer writing onset times than high-frequency ones, this would imply that the first constituent is effectively employed in retrieving the whole compound. Similarly, if low whole-word frequency compounds elicit longer writing onset times than high-frequency ones, it would imply that the whole-word form is effectively employed in retrieving the compound. In visual comprehension, both types of manipulation might exert an effect (e.g., Bertram & Hyönä, 2003; Pollatsek et al., 2000; Burani & Caramazza, 1987; Taft, 1979) or, under some circumstances, only one of them does (e.g., Bertram et al., 2000; Bradley, 1980; Colé et al., 1989; Vannest & Boland, 1999).

5. Frequency as a diagnostic tool Mean 1 st -constituent frequency a 36.138.678.75.7 Mean 2 nd -constituent frequency a 53.173.939.138.9 Mean surface frequency a 8.80.63.84.3 Mean word length b 10.811.011.511.4 Mean 1 st constituent length b 5.65.25.66.3 Mean bigram frequency c 6.37.16.37.7 Mean initial trigram fequency c 0.480.710.650.76 Mean final trigram frequency c 0.831.290.771.05 Mean naming score0.980.910.950.91 Mean visual complexity rating d 4.204.324.064.14 High frequency whole word form Low frequency whole word form High frequency first constituent Low frequency first constituent Mean typicality rating d 2.882.552.892.66 N = 10

6. Method Participants. 18 students of the University of Turku participated in the experiment. All were native speakers of Finnish, and had normal or corrected-to-normal vision. Materials. Before the experiment proper, we conducted a paper-and-pencil experiment to assure that the pictures would elicit the intended compounds. In this pre-experiment 15 native Finnish students participated. They had to write down the name of thirty-one target pictures and 19 filler pictures. Next to naming, they had to rate the visual complexity and typicality of the pictures on a scale from 1 to 5. For the experiment proper, only pictures were included that elicited at least 73.3 of the time the intended compound (average 94.4%). This yielded twenty-six pictures eliciting longish two-noun compounds (on average 11 characters). The lexical statistics of these compounds were based on an unpublished computerized newspaper corpus of 22.7 million word forms, assessed with the help of the WordMill database program of Laine and Virtanen (1999). In the experiment proper, two frequency contrasts were built in. The first contrast was based on the frequency of the whole word form. One condition included 10 items with a relatively high frequency whole word form (8.8 per million) and another condition included 10 items with a low frequency whole word form (0.6 per million). The two conditions were matched on all kind of relevant lexical factors. Since the experiment was a picture elicitation experiment, we matched the conditions on typicality and visual complexity as well. The second contrast was based on the frequency of the first constituent. Hence one condition included 10 compounds with a high frequency first constituent (78.7 per million) and the other included 10 compounds with a relatively low frequency first constituent (5.7 per million). Again, the conditions were matched on all kind of relevant lexical and visual factors. Note that the strength of the frequency manipulation was approximately equal for both contrasts (for whole word manipulation, the factor is 8.8/0.6 ≈ 15, for the first constituent manipulation 78.7/5.7 ≈ 14).

7. Lexicalization in spoken word production lemma level lentokone morpheme level lento kone tuhkakuppi tuhka kuppi ROUTE 1 ROUTE 2 WRITING ONSET TIME (WOT): HF 1406 ms vs LF 1723 ms, p’s ROUTE 1 conceptual level HF LF Mean surface frequency a 8.80.6 Mean 1 st -constituent frequency a 36.138.6

8. Lexicalization in spoken word production lemma level taskulamppu morpheme level tasku lamppu silitysrauta silitys rauta ROUTE 2 WRITING ONSET TIME (WOT): HF 1588 ms vs LF 1600 ms, n.s. ROUTE 3 conceptual level HFLF Mean surface frequency a 3.84.3 Mean 1 st -constituent frequency a 78.75.7

9. Lexicalization in spoken word production lemma level lentokonetuhkakuppi ROUTE 1 conceptual level In simple words, writers retrieve the whole word lemma, before they start to write and there is no morphological involvement

10. Lexicalization in spoken word production lemma level lentokonetuhkakuppi ROUTE 1 conceptual level Question: Is there morphological involvement at a later stage? And how can we find out? lexeme level /lentokone//tuhkakuppi/ morpheme: lento kone syllable: len kotone

11. Lexicalization in spoken word production Question: Is there morphological involvement at a later stage? And how can we find out? l e n t o k o n e INTER-KEYSTROKE INTERVALS (IKI) 12345678 Transition time of adjacent letters without sublexical boundary:224 ms (1: le; 2: en; 4: to; 6:ko; 8: ne Transition time of adjacent letters with syllable boundary:276 ms (3: nt; 7: on) Transition time of adjacent letters with syllable/morpheme boundary:383 ms (5: ok)

12. Lexicalization in spoken word production In sum, when we start to write out a word, we have the lemma readily retrieved, plus the letters of the first syllable, but subsequent individual letters are retrieved (or perhaps reretrieved at the sublexical boundaries) lentokone len WRITING RETRIEVAL l e n WRITING RETRIEVAL WRITING RETRIEVAL WRITING RETRIEVAL to ko ne linguistic planning motor program

13. Conclusions Initial retrieving of compound words is based on the whole concept Before starting to write, the motor program is instructed to write the first syllable Subsequent instructions to the motor program for writing graphemes are given at sublexical boundaries In general, linguistic planning and motor program are much more intertwined than when might think