Confrontation Naming of Environmental Sounds

The development of a set of everyday, nonverbal, digitized sounds for use in auditory confrontation naming applications is described. Normative data are reported for 120 sounds of varying lengths representing a wide variety of acoustic events such as sounds produced by animals, people, musical instruments, tools, signals, and liquids. In Study 1, criteria for scoring naming accuracy were developed and rating data were gathered on degree of confidence in sound identification and the perceived familiarity, complexity, and pleasantness of the sounds. In Study 2, the previously developed criteria for scoring naming accuracy were applied to the naming responses of a new sample of subjects, and oral naming times were measured. In Study 3 data were gathered on how subjects categorized the sounds: In the first categorization task – free classification – subjects generated category descriptions for the sounds; in the second task – constrained classification – a different sample of subjects selected the most appropriate category label for each sound from a list of 27 labels generated in the first task. Tables are provided in which the 120 stimuli are sorted by familiarity, complexity, pleasantness, duration, naming accuracy, speed of identification, and category placement. The WAV sound files are freely available to researchers and clinicians via a sound archive on the World Wide Web; the URL is http://www.cofc.edu/~marcellm/confront.htm.

A testing procedure commonly used in psychology is visual confrontation naming, a task in which subjects are asked to identify simple line drawings or pictures on demand. Cognitive psychologists have used visual confrontation naming to explore the organization of information in, and the retrieval of information from semantic memory, developmental psychologists have used visual confrontation naming to investigate age-based changes in word-finding ability, and clinical neuropsychologists have used visual confrontation naming to screen for word-finding difficulty in neurologically impaired populations. A review of studies using the visual confrontation naming procedure reveals that a wealth of ''raw materials'' exists for conducting such investigations. In the realm of cognitive psychology, the most comprehensive resource is the Snodgrass and Vanderwart (1980) set of 260 normed drawings; these are widely and flexibly used in experiments on naming, categorizing, and processing of pictures versus words. For example, Mitchell (1989) used 96 pictures from the Snodgrass and Vanderwart (1980) set to study developmental changes in naming processes of young and old adults, and Wingfield, Goodglass, and Smith (1990) used 24 Snodgrass and Vanderwart pictures rated high in name agreement to investigate, in aphasic subjects, naming facilitation via presentation of timegated portions of the picture's spoken name. An abundance of normative material also exists for studies of picture naming in the realm of clinical neuropsychology. Several standardized neuro-psychology tests, such as the popular 60-item Boston Naming Test (Kaplan, Goodglass, & Weintraub, 1983) and the 30-item Visual Naming Test of the Multilingual Aphasia Examination (Benton & Hamsher, 1989), provide line drawings normed for use in visual confrontation naming. Such normed picture-naming materials have revealed declines in the naming ability of normal, healthy adults (e.g., Albert, Heller, & Milberg, 1988) and are an integral part of neuropsychological language assessments. Responses may reveal evidence of word retrieval failure, which is often an early marker of dementia or neurological insult (van Gorp, Satz, Kiersch, & Henry, 1986). Furthermore, picture naming is the standard technique for investigating anomia and less severe word-finding problems that are present in most types of aphasia (Benson, 1985;Benton & Hamsher, 1989;Goodglass, 1993;Kay & Franklin, 1995;Miceli, Giustolisi, & Caramazza, 1991;Spreen & Strauss, 1991) as well as in neurological conditions such as Alzheimer's disease (Bayles, Caffrey, Tomoeda, & Trosset, 1990), Huntington's disease (Wallesch & Fehrenbach, 1988), multiple sclerosis (Beatty, Monson, & Goodkin, 1989), Parkinson's disease (Frank, McDade, & Scott, 1996), and stroke (Margolin, Pate, Friedrich, & Elia, 1990).
Common to each of these studies is a nearly universal reliance on a single core task, visual confrontation naming. The practice of using pictures to study naming processes has been encouraged by the excellent quality, careful standardization, and easy availability of several sets of normed pictorial stimuli. It could be argued, however, that overreliance on one task to understand a complex, multifaceted process like naming may ultimately limit the external validity of naming theories as well as our ability to probe the extent of naming difficulties. For instance, if pictures, words, and sounds map onto functionally independent representational systems (cf. Paivio, 1986;Shallice, 1988;Thompson & Paivio, 1994), then it is possible that observed declines in naming ability in older adults -based as they are on picture-naming performanceprimarily reflect changes in visual-structural processes rather than generalized changes in semantic or phonological access.
Although a few investigators have explored the ability to name objects by touch or to name concepts from verbal descriptions, little systematic attention has been given to exploring the other major channel for nonverbally accessing concepts in semantic memory-the auditory modality. As Ballas (1993) noted, ''Few details are known about how we identify and perceive everyday sounds. This is surprising, given the ubiquitous presence of these sounds and their important functional role....There is hardly any theory on this topic, and most of the research has focused on a limited set of sounds,' ' (p. 250). Several have speculated about this lack of progress in the study of nonverbal auditory cognition, citing such factors as the dominance of vision in information processing (Posner, Nissen, & Klein, 1976), the overwhelming emphasis in audition on the study of speech communication (McAdams & Bigand, 1993), and the technical difficulties of working with sound, which, until the recent advent of powerful personal computers, required expensive equipment for editing, splicing, noise filtering, and so on (Luce, 1993).
Studies of nonverbal environmental sounds by cognitive psychologists have focused on topics such as the psychoacoustics of specific events (e.g., the spectral properties of hands clapping or a bottle breaking), the localization of environmental sound sources (e.g., spatial orientation to sounds of differing pitches), the detection of unique acoustic events (e.g., vigilance monitoring of underwater sounds), and the remembering of sequences of environmental sounds (e.g., retention of order information in free recall of sounds vs words). We are aware of four studies in the area of cognitive psychology that have generated normative data on the naming of nonverbal sounds. Ballas (1993) explored how several factors -acoustic, ecological, frequency, perceptual, and cognitive -influence the identification of 41 brief sounds by college students, and Fabiani, Kazmerski, Cycowicz, and Friedman (1996) gathered name agreement and conceptual agreement data on 100 brief sounds in separate samples of children (5-6, 9-11, and 14-16 years), young adults (19-34 years), and old adults (61-88 years and 54-80 years, the latter with Alzheimer's disease). These studies yielded cognitive and developmental insights about sound identification, generated excellent normative data on sets of short sounds, and provided useful stimuli for the investigation of issues such as the role of acoustic factors in sound identification and the electrophysiological study of novel sound recognition. However, the brief duration (under 625 or 400 ms, respectively) and generally low identifiability of the stimuli may limit their usefulness in some situations. As Fabiani et al. (1996, p. 473) noted, the use of short stimuli in sound naming tasks is perhaps more equivalent to naming picture fragments than naming whole pictures. A third study (Chiu & Schacter, 1995) gathered identification data from college students on 1 and 5 s versions of 24 environmental sounds, and a fourth study (Thompson & Paivio, 1994) gathered name agreement and identification latency data on 20 sounds. Limited normative data were gathered on these small sets of sounds (the studies focused on broader cognitive issues and not on the sounds themselves), and the latter study provided few details about the normative sample and procedures. Although two other studies examined the question of whether individuals can accurately name environmental sounds (Lass, Eastham, Parrish, Scherbick, and Ralph (1982) used multiple versions of 14 different sounds, and Van Derveer (1979) (unpublished dissertation) used a set of 30 sounds), their usefulness is limited by insufficient descriptive information about the individual stimuli, the naming responses that they elicited, the manner in which they were scored, and/or their current availability.
Research on sound identification by clinical neuropsychologists has consisted primarily of case studies or small-group studies of braindamaged patients screened for auditory agnosia via sound recognition tasks in which the patient attempts to point to a picture that matches the meaning of the sound (e.g., Coslett, Brashear, & Heilman, 1984;Nagafuchi, Iinuma, Yamamoto, & Kitahara, 1993;Schnider, Benson, Alexander, & Schnider-Klaus, 1994;Stein & Curry, 1968;Van Lancker (et al.) 1988). A smaller group of clinical studies has used naming of familiar sounds to examine auditory perception (e.g., Eustache, Lechevalier, Viader, & Lambert, 1990), anomia (e.g., Goodglass, 1980), and the expressive language ability of individuals with Down syndrome (Marcell, Busby, Mansker, & Whelan, 1998). However, in both kinds of studies (sound recognition and sound naming), the small sets of informal stimuli are of unknown origin and varying quality, unavailable for widespread use, and have no normative data collected on them. Consequently, it is difficult to make comparisons across studies that may have used very different operationalizations of stimuli (e.g., the ''baby cry'' sounds of two studies might actually be very different-sounding stimuli that elicit different responses), and it is also unclear how normal individuals would have responded to the stimuli in terms of ease of naming, familiarity, and so on. We are aware of only one clinically oriented study that has generated normative data on the naming of nonverbal sounds. Benton (1963, described by Spreen &Strauss, 1991) developed the Sound Recognition Test, which has primarily been used by researchers in a four-choice picture-pointing recognition format (Varney, 1980;Varney & Damasio, 1986), although scoring guidelines and normative data are provided for use of the test in an oral naming format. As a tool for examining naming, its strengths include uniqueness (we have been unable to locate any other such published collection of average-length, everyday sounds) and its use of very familiar sounds (normal adults typically score perfectly). Weaknesses include the small number of stimuli (there are two equivalent forms, each consisting of only 13 sounds), insufficient information about the normative sample, and an absence of normative data on both the nameability of the individual stimuli and sound characteristics (like complexity or familiarity) that might influence their identifiability.
To summarize, a wealth of materials is available for studying the naming of pictures, but fewer resources are available for studying the naming of nonverbal sounds. Although the naming of environmental sounds is theoretically relevant to the developmental study of changes in naming ability with normal aging, the experi-1. Unlike Van Derveer (1979) and Ballas and Howard (1987), and largely because of a pragmatic desire to enlarge our stimulus set, we broadened the concept of ''environmental sound'' to include brief snippets of music. Our musical instrument sounds (e.g., accordion, piano) do not appear to function in the traditional artistic sense of music; they function, instead, as brief solo instrument segments that are much more likely to prompt a source identification response (e.g., ''flute'') than an aesthetic response (e.g., ''beautiful crescendo!''). mental study of semantic memory, and the clinical study of word retrieval problems, natural sounds have seldom been used in these efforts, in large part because until recently there has been no normed set of sound stimuli available for use. The primary goal of this project was to create a relatively large set of everyday sounds that can be flexibly adapted for use in clinical and experimental neuropsychological work on sound identification and naming. Our project supplements the work of Ballas (1993) and Fabiani et al. (1996), with the primary difference being that the lengths of our sounds vary with the sound source and the event. Our model for this effort was the Snodgrass and Vanderwart (1980) picture set which, although originally developed for cognitive psychology applications, has also received widespread developmental and clinical use.

Selection and Editing of the Sounds
We conducted a literature review of experimental and clinical studies that used nonverbal sounds and from this developed a list of approximately 80 sounds previously used in research. To this list we added another 40 sounds derived from other sources (informal time samples of our own everyday activities, literature from the heyday of radio sound effects, and serendipitous ''finds'' from searching through sound effects libraries). Digitized stimuli matching the descriptions of these sounds were culled from four CD-ROM royalty-free sound effects libraries, public domain Internet archives, and live recordings. Guidelines for sound selection included clarity, realism, and potential identifiability when presented alone without supporting context. Our initial set of sounds represented a wide variety of different acoustic events such as sounds produced by animals (e.g., cow, dog), people (e.g., laughing, yawning), musical instruments (e.g., piano, trumpet), tools (e.g., hammering, sawing), transportation (e.g., car, airplane), signals (e.g., telephone, doorbell), liquids (e.g., water dripping, ocean waves), and so on. Following Van Derveer (1979), we consider each of these to be an ''environmental sound'' that may be defined as a non-speech sound representing ''...any potentially audible acoustic event which is caused by motions in the human environment,'' (p. 16). 1 As she noted, the physical sources of these sounds can be animate (e.g., yawn, cat meow) or inanimate (e.g., water bubbling, motorcycle), natural (e.g., rain, wind) or artificial (e.g., boat horn, Velcro). Environmental sounds are typically more complex than the usual acoustic stimuli used in laboratories (e.g., pure tones) and are ''meaningful, in the sense that they specify events in the environment,'' (p. 17). They are useful in everyday life in a number of ways, such as warning of danger (e.g., siren), signaling presence (e.g., rattlesnake), denoting correct (e.g., clicking of a stapler) or incorrect (e.g., water dripping) functioning of devices, indicating food crispness (e.g., crunch of celery), locating and orienting to an event (e.g, an explosion to the right), monitoring change in status (e.g., chiming of a cuckoo clock), communicating information about emotional (e.g., scream) or physical (e.g., a burp) state, and so on. Unlike most static pictures, environmental sounds are dynamic in that they convey action and movement-related information -''news that something is happening,'' (Jenkins, 1985, p. 117).
Although a few sounds were used with no changes from the original samples, most were edited; examples of editing changes included reducing duration (e.g., cutting a portion from the middle of a lengthy babbling brook sound), increasing duration (e.g., pasting an additional Ping-Pong hit-and-return sequence), reducing or increasing volume, removing extraneous silences and noises, and applying fade-in and fade-out algorithms to the beginnings and ends of the sounds. An important editing decision was to allow the sounds to vary in length. We take an ecological view of auditory perception (e.g., Jenkins, 1985;McAdams, 1993) and treat acoustic stimuli as complex, dynamic, and informative events with different inherent temporal patterns, ranging from very brief events like the cracking of a whip to lengthier events like the tinkling of wind chimes. Our guideline was to edit each sound to a duration that we believed allowed the ''sound event'' or ''auditory object'' (Wightman & Jenison, 1995) to unfold naturally (cf., Port, Cummins, & McAuley, 1995); this was clearly more of an artistic than empirical endeavor. The sounds were saved as 16-bit .WAV files at a sampling rate of 22,050 Hz.

Common Methods
All participants were undergraduates at the College of Charleston, a liberal arts college in South Carolina that recruits students primarily from the southeast region of the US. At the time of this study, the College's undergraduate population was largely female (62% female, 38% male), white (88% Caucasian, 8% African-American, 4% other), and young (51% 18-20 years, 32% 21-24 years, and 17% > 25 years).
In each of the three studies subjects listened to random orderings of experimental sounds and performed tasks involving either identification, attribute rating, and/or categorization. The sounds were presented free-field via MEL2 software programs (Schneider, 1995) over Pentium computers equipped with stereo speakers. Testing was conducted individually when vocal reaction times were measured and in small groups when written responses were gathered. Participants were initially presented a randomly selected subset of 15-20 experimental sounds in order to provide an idea of the range of different stimuli to be encountered (Snodgrass & Vanderwart, 1980). Practice trials were then presented to ensure familiarity with the task. The typical testing session lasted 45 minutes, with rest breaks taken after 1/3 and 2/3 of the stimuli had been presented. Sounds were presented at a comfortable, preset loudness estab-lished through pilot testing. Each study employed written informed consent and debriefing procedures and conformed with ethical guidelines of the American Psychological Association.

STUDY 1: IDENTIFICATION OF SOUNDS AND RATING OF SOUND CHARACTERISTICS
The purpose of this study was to collect normative data on a large set of everyday sounds, and the primary goal was to gather a corpus of naming responses for use in developing scoring criteria. We recorded the descriptions spontaneously used by subjects to identify the sounds and then used those responses to develop guidelines for scoring naming accuracy. We followed Van Derveer's (1979) nondirective instructional guidelines and did not tell participants, for instance, whether they should describe the sound's perceptual qualities (e.g., ''rapid, lengthy, abrasive sound''), the agent or source of the sound (e.g., ''a person''), the action or event depicted (e.g., ''brushing''), or the recipient of the action (''teeth''). Like Van Derveer, we were simply interested in determining how subjects would describe a sound, and then in using their descriptions to determine the level of generality that would be considered a typical or ''accurate'' description of the sound.
The secondary goal of this study was to determine how subjects rate the sounds on three characteristics that have been shown to influence the identifiability of stimuli: (a) Familiarity refers to how usual or common a stimulus is in the subject's realm of experience; it has been described in visual confrontation naming studies as the pictorial equivalent of word frequency (Basso, Capitani, & Laiacona, 1988). The role of familiarity as a property that influences both picture naming (e.g., Snodgrass & Vanderwart, 1980) and sound identification (Ballas, 1993) is well established; it is also frequently used as a characteristic for selecting subsets of stimuli (e.g., Gainotti & Silveri (1996) created matched picture sets of living and non-living objects of high or low familiarity); (b) Complexity refers to 2. Four sounds were removed from the original set of 118 sounds, 6 new sounds were added, 16 sounds were replaced with new exemplars (e.g., the original helicopter sound was dropped and replaced with a new helicopter sound), and 18 sounds were reedited to achieve better clarity. The final outcome was the set of 120 sounds used in all subsequent studies.
3. The instructions used in Studies 1 and 2 for the sound identification portion of the experiment included the following: ''Your task is to identify each sound as quickly and as accurately as you can. Use one or two words to describe what you hear, and write your response in the blank provided (or 'speak into the microphone as soon as you are ready to respond'). You may write your answer (or 'speak into the microphone') as soon as you are ready to respond, even if the sound is still playing. The sound will be played only once, so listen closely. If you do not know what the sound is, try to guess.'' the amount of perceptual richness, detail, or intricacy of a stimulus (after Basso et al., 1988); it is a feature of nonverbal stimuli that influences performance on identification variables such as naming latency and recognition threshold (Snodgrass & Vanderwart, 1980). The importance of having normative data on complexity can be seen, for instance, in Basso et al.'s (1988) study of the relationship between visual complexity and accuracy in naming pictures from different semantic categories; (c) Pleasantness refers to how pleasing or agreeable a stimulus appears to an individual; it has been shown to be an important emotional attribute of everyday and musical sounds (e.g., Ballas, 1993;Bjork, 1985). Unlike the properties of familiarity and complexity, there is little research available on pleasantness as a dimension of picture naming; we selected it, instead, on the basis of its importance in auditory research topics such as the psychological effects of environmental noise (Vos, 1992). Having normative data on the rated pleasantness of a set of sounds might be useful in cognitive psychology applications such as mood induction (e.g., using potentially unpleasant sounds like ''scream,'' ''jackhammer,'' ''mosquito,'' and ''police siren'' to create a subtle negative mood state in the listener, and sounds like ''ocean,'' ''birds chirping,'' ''harp,'' and ''wind chimes'' to create a positive mood state).

Participants
In the initial study, 25 introductory psychology college students (15 females and 10 males, M age = 19.3 years, SD = 1.4 years) were presented 118 stimuli for judgment. In a follow-up study using identical procedures, a new sample of 25 introductory psychology college students (20 females and 5 males, M age = 20.7 years, SD = 2.4 years) were presented 42 sounds for judgment. The purpose of the follow-up study was to gather new data on subjects' judgments of 42 sounds that had either been altered (reedited, replaced) or added to the original set. 2 All data subsequently presented in Study 1 are for the final set of 120 sounds, each of which was identified and rated by a sample of 25 subjects. Subjects received extra credit for their participa-tion and none responded affirmatively to the selfreport question, ''To the best of your knowledge, do you have a hearing loss?''

Procedure
Participants were tested in small groups of 2-6 individuals. Answers were recorded in writing on slips of paper containing a blank for the trial number (trial #1, #2, etc.), a blank for the name of the sound, and four rating scales. Each of the randomly ordered sounds was presented once, and participants were allowed 30 s to complete their identification and rating tasks (although most finished more rapidly). The participant's primary task was to name the stimulus by writing a response on an 8 cm blank line. Participants were given openended instructions (e.g., ''use one or two words to describe what you hear'') and were not asked to describe, for example, the source of the sound (e.g., ''horse''), the type of sound (e.g., ''galloping''), or the sound qualities (''rhythmic, clip-clop sound on hard surface''). 3 After each identification participants responded to the written question, ''How confident are you in your decision?,'' by circling one of the seven choices on a Likert scale ranging from 1 (''Not at all Confident'') to 7 (''Very Confident''). Finally, participants made 7point Likert scale ratings of the sound's perceived familiarity (1 = ''Highly Unfamiliar'' and 7 = ''Highly Familiar''), complexity (1 = ''Very Simple'' and 7 = ''Very Complex''), and pleasantness (1 = ''Very Unpleasant'' and 7 = ''Very Pleasant'').

Guidelines for Scoring Naming Responses
The initial tabulations of subjects' naming responses revealed that although a few of the sounds were accurately and consistently described with a single word (e.g., ''doorbell''), more of the sounds were accurately described with several different yet completely appropriate words (e.g., ''fiddle'' for ''violin'') or phrases (e.g., ''balling up paper'' for ''crumpling paper''). Thus, it became critical to develop detailed guidelines for evaluating multiple responses to these stimuli, and we did so initially by using the corpus of naming responses to guide us in establishing scoring criteria. For example, tabulation of responses to the baby crying stimulus suggested that a correct description should include both the words ''baby'' and ''crying'' because all 25 subjects spontaneously did so. Likewise, only one subject's response to the sound of a man burping included a reference to the agent that produced the sound; thus, a correct identification in this case needed to refer only to the action itself (the root word ''burp'').
In addition to developing scoring guidelines based on the most frequent responses of participants, we also considered, as did Van Derveer (1979), whether the non-modal responses of participants provided adequate descriptions of the sound events. For instance, we found that although ''truck'' was the most frequent response given to the sound thus labeled, some participants gave a completely acceptable description of this sound at the same conceptual level using the word ''bus.'' Furthermore, we found that even though some non-modal responses did not fall into the dominant conceptual category, they were nevertheless precise alternative descriptions of the acoustic stimulus. For instance, we found that even though ''crickets'' was the most frequent response given to the sound thus labeled, the sound could also be accurately described as that of young birds chirping. In summary, the non-modal description of a sound was judged as correct when the response was one of the following: (1) A synonym for the sound label (e.g., ''teapot'' for ''tea kettle;'' ''mule'' or ''ass'' for ''donkey''). (Unabridged dictionaries and thesauri were used to determine, for instance, that ''pipes'' was an unfamiliar yet acceptable synonym for ''bagpipes,'' and ''chimes'' an acceptable descriptor for ''church bells.'') (2) A description that accurately captured the meaning of the sound source or the conceptual nature of the sound (e.g., ''bomb'' or ''cannon'' for ''explosion;'' ''bouncing ball'' for ''basketball'').
(5) A word or phrase with a different grammatical ending but the correct root word (e.g., a plural version of a singular sound source such as ''ducks'' for ''duck;'' a variation in tense such as ''a car crashed'' for ''cars crashing;'' other root-preserving grammatical modifications such as ''bag piper'' for ''bagpipes'').
(7) An acceptable label given as the second of two different responses (e.g., ''either drums or bongos'' for ''bongos'') [We found in subsequent research with oral responses that the second response was typically the desired response, given usually as a self-correction (e.g., ''Sigh-no, yawn'') or a sharpening (e.g., ''water running-it's a river'') of the first response.] (8) An unanticipated description that was subsequently judged as an acoustically precise alternative interpretation by the unanimous judgments of three independent listeners (e.g., ''fish tank air pump'' for ''water bubbling;'' ''metronome'' for ''clock ticking'').
A sound description was judged as incorrect when the response was one of the following: (1) An inaccurate description of the sound (e.g., ''xylophone'' for ''harp;'' ''vacuum cleaner'' for ''truck''), no response, or a ''don't know'' type of response.
Specific scoring guidelines for determining the accuracy of naming responses to each of the 120 sounds are described in Table 1. The ''Sound Label'' column lists, in alphabetical order, the modal label-the most common descriptor-used to identify a stimulus. For example, the label for the sound of an elephant trumpeting is listed simply as ''elephant,'' and the label for the sound of a door slamming shut is listed as ''door closing,'' because these were the most frequent responses. Stimuli that were inaccurately identified by most subjects are listed either by the most frequently used correct sound label (e.g., ''Velcro'') or by the sound label associated with the original sound recording (e.g., ''can crush''). As in standardized picture-naming tests, each response is scored as correct or incorrect and, where needed, examples are provided in the table of correct and incorrect responses.
Applying the specific scoring guidelines of Table 1 to subjects' naming responses revealed that a sound, on the average, was accurately named by 80.97% of the subjects (SD = 23.94%) with a high level of confidence (M = 5.95, SD = 1.09). As expected, subjects expressed higher degrees of confidence when accurately naming sounds, r(118) = .84, p < .001. We found, as did Van Derveer (1979), that subjects rarely mentioned the agent of a human-produced sound. That is, although many of the sounds were clearly produced by the actions of people (e.g., whistling, can opening, burp, violin, gunshots, laughing, etc.), the majority of subjects mentioned the actor in their descriptions of only two of the sounds -baby crying and child coughing, sounds in which the age information was highly salient; a third sound -''scream'' -was described by 36% of the participants with reference to the actor (most specified a female). Subjects' responses typically focused on either the action/event (e.g., bowling, hammering, shuffling cards) or the inanimate object ( e.g., swords, whip, harmonica) producing the sound. In contrast, all of the animal sounds (e.g., cat, dog barking, chickens, crickets, horse galloping) included reference to the agent of the sound. Participants also tended to generate shorter responses to accurately named sounds (e.g., ''sneeze;'' ''door closing'') and longer responses to inaccurately named sounds (e.g., ''taking top off something'' for ''cork popping;'' ''metal ball rolling'' for ''pinball''), despite the instructions to use only one or two words (cf. Van Derveer, 1979). Finally, we found, as have others (e.g., Ballas & Howard, 1987;Van Derveer, 1979), that participants rarely described the acoustic properties of a stimulus. As Howard and Ballas (1987) put it, ''Simply stated, the recognition of environmental sounds is directed to produce semantic interpretations of the sound,'' (p. 103).

Sound Attribute Ratings and Sound Duration
The sounds are listed in alphabetical order by sound label in the first column of Table 2; mean attribute ratings are listed in the second (familiarity), third (perceptual complexity), and fourth (pleasantness) columns, and stimulus durations are listed in the fifth column. Tables 3-5 list the 120 sounds in order of their rated familiarity, complexity, and pleasantness, respectively, with the highest-rated sounds (i.e., the most familiar, complex, or pleasant sounds) listed at the top of the table.
The distributions of familiarity and complexity ratings strongly paralleled those reported by Snodgrass and Vanderwart (1980) in their normative study of pictures. Familiarity ratings were skewed in the direction of sounds being quite familiar (M = 5.92 on a 7-point scale, SD = 0.98), with the five most familiar sounds being telephone, toilet flushing, birds chirping, brushing teeth, and sneeze (ratings ranged from 7.00 to 6.88) and the five least familiar being can crush, whip, pinball, tearing paper, and sonar (3.44-2.21). Correlational analyses revealed that more familiar sounds tended to be more accurately named, r(118) = .81, p < .001. The very high correlation between familiarity and confidence ratings [r(118) = .96, p < .001] suggests that these two scales were redundant and likely measured the same characteristic. Complexity ratings were compactly and symmetrically dis- Table 1. Specific Scoring Guidelines for 120 Sounds Listed Alphabetically by Sound Label.

Sound label
A correct response must include...

Water draining
Either the object (any liquid) or vehicle (''sink,'' ''tub'') and the act of draining (e.g., ''water going down a drain,'' ''draining a liquid,'' ''tub emptying''); reference to other specific actions with water (e.g., ''flushing,'' ''pouring water,'' ''running water'') not acceptable Water dripping The root ''drip'' or ''drop;'' reference to the liquid alone (e.g., ''water'') or to rain not acceptable Whip The root ''whip;'' reference to a gunshot sound or description of the cracking sound alone not acceptable Whistle (instrument) The root ''whistle'' Whistling (lips) The root ''whistle;'' reference to an event without the root word (e.g., ''calling a dog'') not acceptable Wind The root ''wind'' or any synonym for a strong wind (e.g., ''gale'') Wind chimes Both the roots ''wind'' and ''chime;'' ''bells,'' ''windbells,''or ''chimes'' alone not acceptable Wolf The root word ''wolf'' (e.g., ''werewolf'') or ''coyote;'' reference to ''wild dogs'' or the howling sound itself not acceptable Woodpecker The root ''woodpecker'' Yawning The root ''yawn'' Zipper The root ''zip'' Note. The scoring guidelines listed above were initially developed in Study 1; a few minor revisions were added in Study 2. 1 This sound was actually given a different modal label by participants. The non-modal sound label listed in the table was used for the following reason: (1) The bongos sound was labeled with the root ''bongo'' or ''conga'' (similar-sounding instruments played with the hands) by 38% of the participants and with the root ''drum'' by 48%; however, three independent judges deemed the broader ''drum'' response unacceptable in light of the relative uniqueness of the bongo sound, its clear difference from the other ''prototypical'' drum sound in the set, and the attempts of some subjects to specify a type of drum (e.g., ''tympani'').
(2) The sound of a crow was labeled ''bird'' by 42% of the participants and ''crow'' by 34%; both responses were scored as correct, and the ''crow'' label was kept so as to distinguish it from the other sound involving birds. (3) The sound of a mosquito was labeled ''mosquito'' by 10% of the participants and ''bee'' by 34%; both responses were considered correct acoustic interpretations at the same conceptual level, and the ''mosquito'' label was kept in order to provide the most accurate description of the insect represented in the original sound recording. The bongos and crow sounds were the only examples of sounds not named at a well-agreed upon level of specificity. Generally, we found very close agreement on the appropriate conceptual level at which the sounds were named.
tributed around a mean of 3.26 (SD = .56), with the five sounds rated as most complex being pinball, accordion, sonar, velcro, and can crush (4.92-4.32) and the five rated as simplest being knocking, yawning, doorbell, door closing, and burp (2.24-1.92). Sounds that were perceptually simpler tended to be rated as more familiar, r(118) = -.64, p < .001; this finding parallels Snodgrass and Vanderwart's (1980) finding that visually complex pictures tend to be rated as unfamiliar. Simpler sounds also tended to be named more accurately, r(118) = -.61, p < .001; interestingly, this finding parallels that of Montanes, Goldblum, and Boller (1995), who showed that both healthy older adults and Alzheimer's disease patients show poorer naming of pictures that are high in visual complexity. Pleasantness ratings were widely distributed around a mean of 3.87 (SD = 1.13), with the five most pleasant sounds being ocean, violin, river, saxophone, flute, and piano (the latter two were tied) (6.36-6.08) and the five least pleasant sounds being mosquito, gunshots, jackhammer, scream, and car crash (1.84-1.52). Naming accuracy was not influenced by rated pleasantness of the sounds, r(118) = .14. Furthermore, pleasant- Notes. Standard deviation is in parentheses following the mean. Famil, Complx, Pleas, and Conf ratings were made on 1-7 Likert scales. NAcc scores and Conf ratings were based on the combined scores of 50 subjects from Studies 1 and 2; all other measures were based on the scores of 25 subjects from either Study 1 or 2.
ness correlated only marginally with familiarity, r(118) = .20, p < .05, and not at all with complexity, r(118) = .06, suggesting the relative independence of the pleasantness dimension in rating sound attributes.
The duration of each stimulus is listed in Table 6, with the shortest sounds listed at the top of the table. The overall mean duration of the sounds was 2406 ms (SD = 1306 ms), with the five shortest sounds being a cork popping, can opening, door closing, burp, and pullchain lightswitch (137-447 ms) and the five longest sounds being bowling, truck, saxophone, helicopter, and ocean (4905-6083 ms). As expected on the basis of Ballas and Howard's (1987) and Fabiani et al.'s (1996) research, we found no correlation between the duration of sounds and the accuracy of subjects' naming responses [r(118) = .10]; there was also no correlation between duration and ratings of confidence or familiarity. Duration related weakly but significantly to complexity, r(118) = .24, p < .05, and pleasantness, r (118) = .29, p < .01, with longer sounds tending to be rated as more complex and more pleasant. In summary, Study 1 yielded guidelines for scoring naming responses to a set of 120 computer-presented everyday sounds, as well as normative data on how the sounds are perceived in terms of their familiarity, complexity, and pleasantness. Accuracy in naming sounds was strongly related to higher confidence in naming decisions and higher ratings of familiarity, and sounds rated as more complex tended to be named less accurately, less confidently, and rated as more unfamiliar. There were also weak tendencies for familiar sounds to be rated as more pleasant, and for longer-duration sounds to be rated as more complex and more pleasant. Procedurally, the study demonstrated that participants were able to accomplish several written tasks -naming a sound, rating level of confidence, and rating sound attributes -within a brief period of time.

STUDY 2: ACCURACY AND SPEED OF IDENTIFYING SOUNDS
Study 2 was undertaken to accomplish the following two goals: (a) to apply the scoring criteria developed in Study 1 to a new corpus of naming responses, and (b) to gather normative data on oral response times for naming sounds (cf. Snodgrass & Yuditsky (1996), who measured vocal naming times to the original Snodgrass and Vanderwart pictures). Additionally, a more sensitive measure of hearing ability, the Hearing Screening Inventory (HSI; Coren & Hakstian, 1992), was employed as a screening tool to exclude any potential subjects who might have uncorrected moderate-to-severe hearing loss. The HSI is a brief self-report inventory containing 12 questions about the quality of hearing during everyday situations, such as understanding the words to popular songs. Crossvalidation studies with pure tone audiometry indicate that the HSI is 93.4% accurate in classifying individuals with moderate-to-severe hearing loss (Coren & Hakstian, 1992).

Participants
A new sample of 25 introductory psychology college students (21 females and 4 males, M age = 18.9 years, SD = 1.1 years) participated for extra credit. Their mean HSI score was in the normal range (M = 21.7, SD = 3.9). (HSI scores between 12 and 27 are classified as normal, 28-37 as mild hearing loss, and 38-60 as marked (moderate-tosevere) hearing loss.) No subjects had HSI scores in the moderate-to-severe range, and two subjects had scores of 28 and 31, which placed them in the mild hearing loss range. The mean sound naming accuracy of these two participants (79.6%) was actually slightly better than that of the sample as a whole (74.3%).

Procedure
As in Study 1, participants were presented random orderings of practice and experimental sounds, with each sound presented once for identification. The major procedural differences were as follows: (a) participants were tested individually rather than in groups; (b) responses were spoken into a microphone rather than written; (c) practice sounds were readministered if additional instruction in triggering the voice response key was needed; and (d) attribute ratings (familiarity, etc.) of the sounds were not made. The primary task was to identify each sound as accurately and quickly as possible. The participant initiated the presentation of a sound by pressing a button on a serial response box. The participant was instructed to respond as soon as he or she was ready (even if the sound was still playing), to speak loudly into the microphone, to avoid making extraneous sounds, and to guess whenever necessary. After naming the sound, the participant rated aloud his or her level of confidence in the accuracy of the identification by referring to a 7-point scale taped to the tabletop.
The experimenter transcribed the participant's verbatim identification response and confidence rating; vocal response time (RT) was measured by computer from the onset of the stimulus to the onset of the oral response. On trials in which a noise (e.g., subject clearing throat) rather than the subject's identification response triggered the microphone, the circumstance was noted and the RT discarded (this happened only once during the 3,000 experimental trials). On trials in which the subject's response was of insufficient volume to trigger the voice key, the circumstance was noted and the RT discarded. This occurred on 129 (4.3%) of the experimental trials and appeared to occur most frequently with sounds that were more difficult to identify (e.g., can opening) and sounds whose labels began with soft phonemes (e.g., sheep).

Naming Accuracy and Confidence Ratings
Accuracy of the spoken naming responses was determined independently by two judges using the scoring criteria developed in Study 1. Of the 3000 trials (120 experimental stimuli × 25 participants), there were only 24 scoring disagreements (representing 0.8% of the data); disagreements that were not mistakes were used to refine and reword scoring criteria. Mean naming accuracy was 74.27% (SD = 27.65%), similar to that obtained in Study 1. The slightly lower (6.7%) accuracy score of Study 2 may have reflected differences between the speeded oral response procedure of Study 2 and the more relaxed written response procedure of Study 1. The mean confidence rating in Study 2 was 5.79 (SD = 1.16), quite similar to that obtained in Study 1 (M = 5.95, SD = 1.09). As before, naming accuracy correlated highly with confidence ratings, r(118) = .85, p < .001 (the correlation was .84 in Study 1).
Naming accuracy and confidence rating data obtained in Study 2 for the 120 sounds were averaged with the same measures taken in Study 1; these data are listed in the sixth and seventh columns of Table 2, and the sounds are listed in descending order of naming accuracy in Table 7. The range of accuracies confirms the availability of a large set of nonverbal everyday sounds that are identifiable by the normal population and potentially valuable in establishing baseline abilities for studies of auditory confrontation naming. There is a large subset of 62 stimuli (sounds 1-62, Table 7) nameable at accuracy levels of 90% or higher (the choice of a 90% identifiability level is based on guidelines employed in picture-naming research (e.g., Stimley & Noll, 1991)). This high level of accuracy ensures their easy recognizability by the normal population and, thus, their clinical utility for assessing anomia and their practical utility for probing semantic memory concepts. The subset of 62 easy-to-name sounds more than doubles the number of such sounds referred to in published studies of sound identification. It is interesting to note that most of these sounds were identified with a multiplicity of responses (e.g., ''birds,'' ''bird whistling,'' ''singing birds,'' ''birds chirping'') rather than a single common label. When the criterion of an exact match between a response and the modal sound label was applied, only seven sounds emerged as having been given precisely the same label by over 90% of the participants (e.g., in this tabulation, a response like ''violins'' was not considered a match for ''violin''). The seven sounds named with highly reliable labels were doorbell (100%), cash register (94%), baby crying (92%), harmonica (92%), helicopter (90%), motorcycle (90%), and owl (90%).
The sound set also includes subsets of 17 mildly challenging sounds (sounds 63-79, Table  7) nameable at accuracy levels of 75-89% and 23 moderately challenging sounds (sounds 80-102, Table 7) nameable at accuracy levels of 50-74%. These sounds should provide an excellent range of difficulty for exploring word-finding problems in older adults, and will be particularly useful in providing opportunities for eliciting naming responses that are within the correct semantic category but are either imprecise or not at the basic level of classification (such as ''car'' for the sound of a large truck or ''game'' for the sound of Ping-Pong). This represents a larger pool of mildly-to-moderately difficult items than is available in the 60-picture Boston Naming Test -and without the vocabulary and cultural confounds engendered by BNT items like ''trellis'' and ''yoke'' (Hawkins et al., 1993).
Finally, the sound set includes 18 strongly challenging sounds (sounds 103-120, Table 7) nameable at levels of accuracy under 50%. This subset of difficult-to-identify stimuli may be useful, for instance, in studies exploring individual differences in sound identification (e.g., do young adults with high IQs identify these sounds more accurately and rapidly?) or the role of identification difficulty in recognition memory (e.g., do people show poorer long-term memory for sounds that are hard to identify?). It is interesting to note that at least 10 sounds (can crush, cork popping, frying food, jackhammer, monkey, pullchain lightswitch, seal, stapler, Velcro, woodpecker) -mostly from the subset of difficult-to-identify sounds -appear to be inconsis- tently named because of their acoustic ambiguity and potential for multiple interpretations by the casual listener. Some subjects, for instance, gave the response ''rain'' to the sound of frying food (note, however, that such an interpretation, although superficially similar to the actual stimulus, was not judged as an acoustically precise interpretation of the sound). Although not the primary focus of this project, these sounds could be useful in cognitive psychology experiments, such as Ballas and Howard's (1987) investigation of parallels between speech homonyms and ambiguous ''sound homonyms'' that are inconsistently interpreted in the absence of contextual cues.

Response Time and Stimulus Duration
Participants' mean oral response times to the 120 sounds can be found in column 8 of Table 1; the sounds are listed in ascending order of speed of response in Table 8, with the most rapidly identified stimuli listed at the top of the table.
The overall mean RT was 3398 ms (SD = 1103 ms). The five fastest RTs were to the burp, car horn, whistle (instrument), telephone, and dog barking sounds (1511-1664 ms) (interestingly, three of these are signal sounds whose purpose is indeed to provoke a rapid response); the five slowest RTs were to the ocean, blinds, pinball, shuffling cards, and sonar sounds (5573-6643 ms). As expected, subjects who responded more slowly to sounds tended to name the sounds less accurately, r(118) = -.67, p < .001, and to have less confidence in the accuracy of their responses, r(118) = -.67, p < .001. 4. Given the correlation between stimulus duration and the speed of the subject's naming response, it is nevertheless interesting to note that visual inspection of subjects' records indicated that they understood the instructions to respond as quickly as possible, even if it meant responding during the presentation of a sound (e.g., mean RTs to 7 of the 10 longest sounds were shorter than the duration of the sounds themselves). As in Study 1, there was no correlation between the duration of the sounds and either the accuracy of subjects' naming responses, r (118) = -.01, or their level of confidence in the naming responses, r(118) = .13. However, duration did correlate with RT, r(118) = .56, p < .001: When presented longer sounds, participants began their naming responses more slowly. This is to be expected given that sounds were edited to the minimal length that allowed the ''sound event'' to unfold (thus, longer events should require more listening before responding). 4 In summary, Study 2 revealed that the previously developed criteria for scoring sound naming accuracy can be employed by independent judges at a high level of agreement. The current set of 120 stimuli contains 62 items that are easy to identify, 40 that are mildly-to-moderately dif-ficult, and 18 that are hard to identify; furthermore, 10 items from the last two groups tended to elicit multiple interpretations because of their acoustic similarity to other sounds. Finally, correlational analyses revealed the following relationships: (a) accurately named sounds were rated more confidently; (b) rapidly named sounds tended to be named more accurately and with greater confidence; and (c) longer sounds tended to be named more slowly.

STUDY 3: CATEGORIZATION OF SOUNDS
The importance of categories in simplifying and structuring human information processing is clear, and several taxonomies have been offered for describing the structure of mental categories. For example, Rosch's (1975Rosch's ( , 1978 pioneering work introduced the notions of superordinate, base level, and subordinate categories, with natural objects initially classified as base-level category members, and category members differing in their degree of prototypicality. Research in cognitive psychology using a variety of techniques (e.g., priming, rating, visual confrontation naming) strongly suggests that knowledge of objects, at least as assessed through pictures, is organized within long-term memory by natural, base-level semantic categories rather than, for instance, categories based on visual-perceptual features like color or size.
A considerable amount of normative research has established guidelines for how people categorize words (e.g., Battig & Montague, 1969). These word category norms have also been used by other researchers to categorize the naming of pictures (e.g., Snodgrass & Vanderwart, 1980). In the present project, however, we have not found word-based category norms to be especially helpful in determining appropriate category placements for the names of sounds. Many of the sounds in Studies 1 and 2 do not appear in lists of word or picture category norms, and, upon reflection, it seems unwise to try to categorize sounds with labels that were obtained through judging words. We have located a few acoustic categorization schemes for everyday sounds devised by other researchers, but have found the number and types of categories proposed to be either limited, artificial, or difficult to apply. For instance, Gaver (1993) noted that sound effects libraries typically list sounds by the context in which they are heard (e.g., traffic, office, household) -a system which has the disadvantage of not using mutually exclusive categories (e.g., is a telephone ringing sound properly classified as an office or household item?). He proposed, as a beginning, a physics-based hierarchical scheme for classifying sounds by type of interacting materials (vibrating objects, aerodynamic sounds, and liquid sounds), with nested subcategories of more specific types of sounds (e.g., two types of liquid sounds: dripping and splashing). Other researchers have proposed different classification schemes (e.g., Porteous and Mastin's (1985) division of the urban soundscape into natural, human, activity, indicator, and neighbor sounds). The most common approach, however, has been for investigators to create their own pragmatic, ad hoc categorizations of everyday sounds (e.g., Ballas' (1993) description of sound types as signals, modulated noise, multiple mechanical transients, discrete impacts, and water sounds; Lass et al.'s (1982) classification of sounds as animal, inanimate, musical, and human; van Lancker et al.'s (1988) categorization of sounds as human vocalizations, animal vocalizations, inanimate noises, and event sounds). Because there is no agreed-upon, all-encompassing system of classifying everyday sounds, we decided to take an empirical approach to the problem and simply ask subjects to generate their own classifications of sound stimuli.
In an early pilot effort -before we had edited, digitized sounds available to use as stimuli -we asked nine subjects to read, imagine, and then categorize brief written descriptions of sounds. The subjects had unlimited time to sort the 150 written descriptions into any number of clusters. The results showed that the number of categories generated by subjects ranged from 12 to 28, with a mean of 19. Several categories (e.g., animal sounds, human sounds, musical instruments, tools) resembled the ad hoc categories typically created by researchers, appeared with consistency across subjects, and showed good agreement on the sounds that were members. However, there were also several personal and idiosyncratic categories typically used by one or two individuals (e.g., sounds heard on a New York street corner, loud sounds). Our intention in Studies 3A and 3B was to use the same general approach that we used in the pilot test -ask subjects to categorize sounds -but to do so in a way that yielded better agreement and fewer ''minority'' categories.

STUDY 3A : FREE CLASSIFICATION
The purpose of this study was to develop a list of category labels that could be used to classify the environmental set of 120 sounds.

Participants
Thirty-eight introductory psychology college students (27 females and 11 males; M age = 19.4 years, SD = 2.0 years) participated for extra credit. The mean HSI score was in the normal range (M = 20.1, SD = 4.1). No participants had scores in the moderate-to-severe range, and two participants had scores of 29 and 33, consistent with a mild hearing impairment. The mean naming accuracy of these two subjects (76.7%) was similar to the mean naming accuracy for participants in Studies 1 and 2 (77.6%).

Procedure
Testing was accomplished in small groups of 1-5 individuals. Answers were recorded in writing on slips of paper containing two numbered blanks (each 8 cm long) for the name and category description of the sound. Participants identified and categorized one of two randomly assigned sets of 60 sounds using a modification of McAdams' (1993) 'free classification' procedure. After listening to two repetitions of a sound separated by 2 s, participants were allowed 15 s to write both the name of the sound and a brief description of the category to which the sound belonged. Participants were told that categorization involves placing something with other objects that have similar characteristics and are members of the same group. The category description could be at whatever level of abstraction the subject chose (e.g., one subject might categorize the sound of birds chirping at the basic level of ''bird;'' another might classify it at the superordinate level of ''animal;'' yet another might create a perceptual category of ''high-pitched, whistle-like sounds''). Subjects were encouraged to develop and/or reuse any number of category labels of their own choosing, with the constraint that each sound be given only one ''best'' category label. Subjects were also asked to avoid giving simple associations to the sounds (e.g., to avoid using a category label like ''saddle'' or ''cowboy'' for the sound of a horse neighing).
Once all 60 sounds were identified and categorized, each sound was played again in order to allow subjects to review and, if necessary, change their category descriptions. During the review procedure each sound was presented once with a 5 s interval between stimuli. The entire experimental task was preceded by a practice procedure that required the generation of names and category labels for three pictures (elephant, snake, glass) and four practice sounds (clock ticking, cowbell ringing, horse neighing, squeaky door opening). The naming data gathered for the 60 sounds were not scored and were used only to clarify, if needed, the nature of the proposed category.

RESULTS AND DISCUSSION
Free classification yielded a variety of category descriptions of the sounds. Two judges independently reviewed the descriptions offered by subjects and grouped those descriptions that were judged as essentially equivalent in meaning into the same category (e.g., the category name ''instrument'' was considered equivalent to the modal phrase, ''musical instrument''); disagreements were discussed and resolved. Any category used by over 33% of the sample to classify a sound was culled for later use in Study 3B. There were 23 such categories: 4-legged animal, accident, air transportation, bathroom, bird, farm animal, game/recreation, ground transportation, human, hygiene, insect, kitchen, musical instrument, nature, pet, reptile/amphibian, sickness, signal, sleep, tool, water/liquid, weapon, and weather. Categories that classified only one sound (e.g., the category ''photography'' for the sound of a camera) were not culled because of their limited usefulness (cf. Snodgrass & Vanderwart, 1980). Three additional categories (household, paper, and machine) that did not meet the 33% inclusion criterion were nevertheless culled from subjects' responses for later use in order to provide potential vehicles for classifying hard-to-categorize sounds; furthermore, the category ''other'' was added to the list in order to give future subjects an option to use whenever they are unable to classify a sound. Finally, the wording of a few category labels was altered slightly from the modal labels in order to avoid overlap (e.g., ''aircraft'' and ''transportation'' were changed to ''air transportation'' and ''ground transportation''). In summary, a total of 27 sound category labels were generated in Study 3A; these can be found as the bold items in Table 9.

STUDY 3B: CONSTRAINED CLASSIFI-CATION
The purpose of this study was to determine how a new sample of participants would apply the category labels generated in Study 3A to the collection of 120 experimental sounds. This study employed a ''pure'' classification task in which the stimuli themselves were not overtly named.

Participants
Forty-nine introductory psychology college students (35 females and 14 males; M age = 20.6 years, SD = 3.3 years) participated for extra credit. The mean HSI score was in the normal range (M = 20.8, SD = 4.7). No subjects had scores in the moderate-to-severe range, and five subjects had scores of 28 or higher (M = 30.2, SD = 2.3), consistent with a mild hearing impairment. Because sound-naming data were not gathered in this study, we were unable to compare the sound identification abilities of these five participants with the norms established in earlier studies. However, visual inspection of the data suggested that their categorizations of the sounds did not differ substantially from those of the larger group.

Procedure
Testing was accomplished in small groups of 1-5 individuals. Participants categorized one of two randomly assigned sets of 60 sounds using a constrained classification procedure (McAdams, 1993) in which they listened to two presentations of a stimulus separated by 2 s. They were allowed 15 s to select from a handout the single ''best'' category label for the sound and to write the label on an answer sheet. As before, they were instructed to think about how the sound might be classified with other things that are similar and are members of the same group. Participants were randomly assigned one of two versions of the handout containing the 27 category labels generated in Study 3A. In one version of the handout the labels were listed in alphabetical order and in the other version the labels were listed in reverse alphabetical order. The category label ''Other'' always appeared at the end of the list and its use was discouraged. At the end of the experimental task each sound was re-presented once, at 5 s intervals, to allow participants to review and, if necessary, change their category choice.
Before categorizing the experimental sounds, subjects completed two practice procedures. In the first practice procedure subjects selected category labels for three line drawings (apple, bed, kite) from a handout with 10 picture-oriented category labels like ''clothing,'' ''furniture,'' and ''vegetable.'' In the second practice procedure subjects selected category labels for the four Study 3A practice sounds from the handout of 27 sound category labels. Instructions noted the following points: (a) although several possible category labels could be selected to describe a stimulus, the task was to pick the ''best'' category; (b) category labels could be reused for different stimuli; and (c) all category labels did not have to be used (recall that each subject categorized only half of the 120 sounds).

RESULTS AND DISCUSSION
The results of the constrained classification task are presented in Tables 9 and 10. Table 9 lists the 27 categories, the sounds that were assigned to them, and scores representing the percentage of subjects who agreed on a sound's category placement. For example, 26 of 28 subjects (93%) agreed that the chickens sound was best placed in the farm animal category, a value that represents a good fit between the chickens sound and the farm animal category; the two remaining subjects believed that the chickens sound was best placed in the bird category (7% agreement), a value that represents a poor fit between the chickens sound and the bird category. The ranking of sounds within each category in Table 9 is in order of their category placement score; thus, a sound that was assigned to the same category by all or most subjects will be found near the top of that category list. Sounds within a category that had tied scores are listed alphabetically, and sounds assigned to a category by fewer than 20% of the subjects are not listed in the table. Some of the sounds in Table 9 are listed under more than one category (e.g., 61% of the sub- jects placed the duck sound into the bird category and 36% placed it into the farm animal category, resulting in the duck stimulus being listed twice in the table). We followed Snodgrass and Vanderwart's (1980) procedure for multiple listings of stimuli in different categories: An asterisk next to the listing(s) of the sound with the lowest category placement score indicates that the sound appears with a higher score in another category.
The data in Table 9 indicate that categories differed widely in size, ranging from seven small categories with only two members (accident, air transportation, bathroom, hygiene, insect, pet, and sleep) to five large categories with eight or more members (4-legged animal, human, musical instrument, signal, and tool). Inspection of the data also suggest the following observations: (a) The musical instrument category was both the largest category, with 17 members, as well as a category in which sounds were placed at very high levels of agreement. (b) Almost half of the sounds of another large category, the signal category, were categorized at low levels of agreement (under 50%), with four of the sounds appearing at higher levels of agreement in other categories. (c) The ''other'' category was used sparingly, with its three members possessing low category placement scores. (d) The 10 hard-to-identify, ambiguous sounds that had yielded multiple naming responses in Studies 1 and 2 also tended to yield unusual category placements in this study. For instance, the monkey sound was most frequently placed in the bird category; 29% of the subjects believed the can crush sound to be the sound of a 4-legged animal; 25% categorized the sound of blinds as that of a machine; and over half of the subjects assigned the frying food sound to the weather and water/liquid categories, suggesting a confusion with the sound of rain. (e) Future investigators of sound categorization who wish to reduce the number of categories might consider consolidating some of the smaller categories (e.g., combining the two-member hygiene and bathroom categories; eliminating the twomember sleep category, with its members likely recategorized as human sounds). Table 10 contains a listing of the 120 sounds in alphabetical order, with their category memberships and category placement scores nested underneath. As expected, the ease with which sounds were placed into categories varied widely across stimuli. Overall, 50 sounds were placed with high agreement (90% and above) into categories, 58 were placed with mild-tomoderate levels of agreement (50-89%), and 12 were placed with low levels of agreement (49% and below). Some stimuli -like the sounds of airplane, laughing, and toilet flushing -were classified with perfect agreement by subjects into the categories of air transportation, human, and bathroom, respectively. Other sounds -like can crush, pullchain lightswitch, and zipperwere difficult to categorize, with subjects using numerous category labels at low levels of agreement. Informal inspection of the data suggested that difficult-to-categorize sounds also tended to be difficult-to-name sounds (e.g., the sounds of can crush, cork popping, and pullchain lightswitch showed low levels of agreement in both the naming tasks of Studies 1 and 2 and the constrained categorization task of Study 3B). However, the relationship between ease of naming and ease of categorizing was by no means perfect: Some sounds named at low levels of accuracy in Studies 1 and 2 were nevertheless categorized at high levels of agreement in Study 3B. For example, although only 18% of the subjects correctly named the sonar sound and only 38% correctly named the bongos sound, 86% and 90% agreed on their appropriate placements within the signal and musical instrument categories, respectively. Conversely, some sounds that were very accurately named in Studies 1 and 2 (e.g., cuckoo clock, knocking, and telephone) generated disagreement among subjects as to their best category placement. For instance, although 100% of the subjects agreed on their naming of the telephone sound, there was disagreement as to its most appropriate category placement, with 43% placing it in the household category, 43% in the signal category, and 14% in the machine category.   Note. Sounds are listed in alphabetical order. Category assignments for each sound are listed in descending numerical order (tied category assignments are listed in alphabetical order). Sound stimuli were randomly divided into two sets of 60 for categorization. One set was evaluated by 21 subjects and the other set by 28 subjects. 1 Indicates that the total does not equal 100% due to rounding to the nearest whole number.

GENERAL SUMMARY
The goal of this research report has been to describe the development of a set of 120 digitized environmental sounds to use in sound naming tasks. The sounds vary in duration from 137 ms (cork popping) to 6083 ms (ocean), and normative data are reported on how they are identified (naming accuracy, rated confidence, naming speed), how their attributes (familiarity, complexity, pleasantness) are rated, and how they are categorized (number and types of categories, number of sounds placed into each category, percentage of agreement on sound placement). These normative data should allow researchers maximum flexibility in selecting subsets of sounds for use in particular applications (e.g., subsets of sounds that are easy or difficult to name; subsets of sounds rated low, medium, or high on familiarity). As in recent normative studies of pictures (e.g., Cycowicz, Friedman, & Rothstein, 1997) and sounds (Fabiani et al., 1996), our stimuli are freely available to researchers and clinicians for professional use via download of compressed digital files over the World Wide Web; the URL is http://www.cofc .edu/~marcellm/confront.htm. Because many of the stimuli are edited versions of sound clips that were originally recorded and made commercially available for royalty-free use by sound effects library vendors, use of the sounds is restricted to non-profit scientific and clinical in-vestigations; users are prohibited from collecting and republishing the items as a separate test or as part of a sound effects collection. Once the sound files are obtained from the web site they can be converted via generic sound-editing software to non-WAV formats and incorporated into any of several commercially available multimedia software programs for presentation as a ''slideshow'' of sounds.

Limitations of the Study
It should be remembered that the normative samples in this investigation consisted of young adult college students whose results may or may not generalize to different populations. Researchers should be particularly sensitive to the possibility of unanticipated but correct responses of individuals from different age groups. For instance, although none of the young adults described the helicopter sound as a ''whirlybird,'' the harmonica sound as a ''mouth organ,'' or the accordion sound as a ''concertina,'' these are appropriate descriptions that we anticipated as potential synonyms and included in the scoring guidelines. It is unlikely, however, that we anticipated all such responses. It is also possible that there will be cohort differences in the familiarity of some of the sounds. For instance, we suspect that old adults might actually be more likely than young adults to recognize and name the sounds of a manual typewriter and a 1960s era pinball machine -sounds that were accurately identified by only 26% and 36% of the young adults in the current study. We hope that researchers will use the sound set to establish normative data for populations in which they are interested, and we also hope that researchers will gather additional normative data on different aspects of the sounds, such as their acoustic similarity (cf. Ballas, 1993).
We believe that the responses of the subjects in our investigation provide an excellent depiction of how ''average'' individuals describe the sounds (it was typically the case that the modal sound label derived from subjects' responses closely matched the label given to the original sound file by its creator). It should be remembered, however, that arbitrary decisions made during the development of our scoring guide-lines clearly determined what was considered an ''accurate'' naming response. First, the procedure of using the participants' modal response as the target sound label occasionally yielded unexpected scoring guidelines. For instance, although it seemed reasonable to expect that some subjects listening to the car horn stimulus would respond with a simple description of the source of the sound (''car''), it was actually the case that no subject responded with only the word ''car,'' and all subjects responded with the root word ''horn'' used alone or in combination with ''car.'' Thus, a scoring guideline was established in which a response referring only to the source of the sound was considered incorrect (recall that one goal was to determine the level of specificity at which subjects would name the sound and to accept all synonyms at this level as correct). Second, it should also be remembered that participants' naming responses were gathered using open-ended instructions to name the sounds (not directive instructions to describe the source of the sound, the type of sound, or so on) and were scored using a correct-or-incorrect system. It is possible that what is considered an accurate response could change, for instance, under more directive instructions (e.g., Ballas (1993) asked subjects to use both a noun and a verb to name the sound) or a more lenient scoring system (e.g., Van Derveer (1979) used a 0-, 1-, or 2-point scoring system). We encourage users of the sound set to continue to refine our proposed scoring guidelines for sound naming and to develop alternative scoring systems that might be more appropriate for specific applications. For example, some may find it useful to develop a scheme for distinguishing between semantic, perceptual, and phonological soundnaming errors (cf. Albert et al., 1988;Vitkovitch, Humphreys, & Lloyd-Jones, 1993), or to create separate norms for name agreement and conceptual agreement (e.g., Fabiani et al., 1996).

Possible Uses of the Sound Set
Although the intended use of these environmental sounds is in auditory confrontation naming applications such as the study of word-finding difficulties with special populations, the sounds can be flexibly adapted for a variety of applications in clinical and experimental neuropsychology research on sound identification. Here are a few possibilities: Several studies have reported progressive impairment of visual confrontation naming in Alzheimer's patients (e.g., Bayles et al., 1990;Daum, Riesch, Sartori, & Birbaumer, 1996;Jacobs et. al., 1995;Kirshner, Webb, & Kelly, 1984;Locasio, Growdon, & Corkin, 1995;Pollmann, Haupt, & Kurz, 1995). A variety of analyses (e.g., types of naming errors, consistency of naming performance when retested with repeated items) suggest that many Alzheimer's patients show an actual degradation or loss of semantic information about a concept, not just an inability to retrieve a concept's name (Frank et al., 1996;Henderson, Mack, Freed, Kempler, & Anderson, 1990;Shuttleworth & Huber, 1989). It would seem important, however, to determine if the performance patterns evidenced by Alzheimer's patients are tied to the specific mode (visual) by which they attempt to access the presumably degraded knowledge. Given recent research on the functional independence of visual and auditory representations of central concepts (Thompson & Paivio, 1994) and the theoretical possibility of separate and redundant modality-specific LTM stores (e.g., Kieras, 1978), it would seem useful to determine whether difficulty in naming an object or event depicted pictorially (e.g., a drawing of a typewriter) also occurs when the same concept is portrayed acoustically (as the sound of someone typing). Because of the traditional focus on confrontation naming of pictures, research on normal and impaired word-finding processes has been primarily concerned with the retrieval of object labels (nouns) (McCarthy & Warrington, 1990). Although action labels (verbs) can be depicted in single, static pictures, it is likely that temporally unfolding sounds can depict dynamic processes less ambiguously than pictures (cf. Jenkins, 1985). Thus, development of a set of sounds could broaden the range of to-be-named stimuli to include action labels in addition to object labels, and perhaps facilitate the investigation of classspecific naming difficulties, such as the verbs-worse-than-nouns pattern shown by some demented patients (White-Devine et al., 1996). Are naming difficulties in anomic subjects the same across different categories of sounds? There are detailed cognitive neuropsychological case studies of brain-damaged patients who appear to have highly selective picture-naming impairments. For example, patients have been described with strikingly selective anomias for fruits and vegetables (Farah & Wallace, 1992), animals (Temple, 1986), living beings (Gainotti & Silveri, 1996), foods and living things (Warrington & Shallice, 1984), and faces (Burton & Bruce, 1992). Not only do such mental dissociations suggest support for the existence of functional modular brain organization (Funnell, 1987), but they also provide important constraints for models of knowledge organization. No such category-specific naming impairments, however, have been identified on the basis of sound naming. Perhaps this is due to the rarity of such selective losses, but perhaps it is also due to the unavailability of a sufficiently diverse set of test sounds with multiple exemplars from each category to use in testing such patients. Investigators have reported evidence of modality-specific impairments (e.g., ''optic aphasia'') in naming from vision, touch, and verbal description (e.g., Beauvois, 1982). Such naming disorders are consistent with empirical evidence for the functional independence of visual and auditory representations of central concepts, whether manifested by independent access routes to a common system (e.g., Riddoch & Humphreys, 1987) or separate and redundant modality-specific LTM stores (e.g., Kieras, 1978). Development of a normed set of sounds with a wide range of difficulty is essential if investigations of modality-specific naming impairments are to be extended to the naming of nonverbal acoustic stimuli. Theoretical frameworks for understanding the organization and functioning of semantic long-term memory have been developed largely on the basis of subjects' responses to linguistic and pictorial stimuli. A variety of experiments involving the naming of nonverbal, everyday sounds could be useful in exploring both the generality of effects discovered in studies with words and pictures and the nature of knowledge organization in semantic long-term memory. For example, does the facilitation effect seen in within-modality repetition priming of words and pictures hold for everyday sounds? Sounds ranked high on familiarity and presented in an initial session (e.g., a yawn rapidly masked with noise to prevent explicit identification) could be represented later in their original form; the empirical question is whether the re-presented sounds are named more rapidly than a set of new sounds, matched on familiarity, from the same category (e.g., a cough). Also, will subjects identify a target sound faster after having been primed earlier by a sound from the same semantic category (''dog barking'' as prime sound, ''wolf howling'' as target sound) or by a sound that is from a different semantic category but has similar perceptual characteristics (e.g., ''person whistling'' as prime, ''birds singing'' as target)? Similar studies have recently been performed by Jones (1995, 1996) and Chiu and Schacter (1995); it is hoped that ready access to a large set of normed environmental sounds might encourage more such efforts.